Add 25 LoRA(s): loom-style-anima, krea-2-raw-true-first-person-pov, kreamolino, moody-gal-vivian-krea2-zimage…

.gitattributes

@@ -1462,3 +1462,142 @@ sick-dolls-brigette-overwatch/example_images/03_image.jpeg filter=lfs diff=lfs m
 sick-dolls-brigette-overwatch/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
 sick-dolls-brigette-overwatch/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
 sick-dolls-brigette-overwatch/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+loom-style-anima/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+loom-style-anima/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+loom-style-anima/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+loom-style-anima/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+loom-style-anima/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+loom-style-anima/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krea-2-raw-true-first-person-pov/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krea-2-raw-true-first-person-pov/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krea-2-raw-true-first-person-pov/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krea-2-raw-true-first-person-pov/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krea-2-raw-true-first-person-pov/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krea-2-raw-true-first-person-pov/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreamolino/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreamolino/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreamolino/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreamolino/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreamolino/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreamolino/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+moody-gal-vivian-krea2-zimage/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+moody-gal-vivian-krea2-zimage/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+moody-gal-vivian-krea2-zimage/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+moody-gal-vivian-krea2-zimage/example_images/04_poster.jpg filter=lfs diff=lfs merge=lfs -text
+moody-gal-vivian-krea2-zimage/example_images/04_video.mp4 filter=lfs diff=lfs merge=lfs -text
+moody-gal-vivian-krea2-zimage/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+moody-gal-vivian-krea2-zimage/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+retro-anime-style/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+retro-anime-style/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+retro-anime-style/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+retro-anime-style/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+retro-anime-style/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+retro-anime-style/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+crt-vaporwave/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+crt-vaporwave/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+crt-vaporwave/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+crt-vaporwave/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+crt-vaporwave/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+crt-vaporwave/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+russian-graduation-schooluniform-cosplay/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+russian-graduation-schooluniform-cosplay/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+russian-graduation-schooluniform-cosplay/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+russian-graduation-schooluniform-cosplay/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+russian-graduation-schooluniform-cosplay/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+russian-graduation-schooluniform-cosplay/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pok-mon-retro-anime-style-1997/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pok-mon-retro-anime-style-1997/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pok-mon-retro-anime-style-1997/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pok-mon-retro-anime-style-1997/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pok-mon-retro-anime-style-1997/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pok-mon-retro-anime-style-1997/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+skyrim-style/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+skyrim-style/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+skyrim-style/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+skyrim-style/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+skyrim-style/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+skyrim-style/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+sick-dolls-savannah/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+sick-dolls-savannah/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+sick-dolls-savannah/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+sick-dolls-savannah/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+sick-dolls-savannah/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+sick-dolls-savannah/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+elusarca-s-krea-2-detail-enhancer-edit-lora/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+elusarca-s-krea-2-detail-enhancer-edit-lora/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+elusarca-s-krea-2-detail-enhancer-edit-lora/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+elusarca-s-krea-2-detail-enhancer-edit-lora/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+elusarca-s-krea-2-detail-enhancer-edit-lora/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+elusarca-s-krea-2-detail-enhancer-edit-lora/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/example_images/06_video.mp4 filter=lfs diff=lfs merge=lfs -text
+kreabarbucci/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreabarbucci/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreabarbucci/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreabarbucci/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreabarbucci/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+kreabarbucci/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+nizur-girls-narrow-waist-thick-thighs-pear-shaped-figure/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+nizur-girls-narrow-waist-thick-thighs-pear-shaped-figure/example_images/02_video.mp4 filter=lfs diff=lfs merge=lfs -text
+nizur-girls-narrow-waist-thick-thighs-pear-shaped-figure/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+nizur-girls-narrow-waist-thick-thighs-pear-shaped-figure/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+miai-mini-dress/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+miai-mini-dress/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+miai-mini-dress/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+miai-mini-dress/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+miai-mini-dress/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+miai-mini-dress/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+splashpasto-paint/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+splashpasto-paint/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+splashpasto-paint/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+splashpasto-paint/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+splashpasto-paint/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+splashpasto-paint/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pov-slider/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pov-slider/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pov-slider/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pov-slider/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pov-slider/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+pov-slider/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+jnkku-style-anima-krea2/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+jnkku-style-anima-krea2/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+jnkku-style-anima-krea2/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+jnkku-style-anima-krea2/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+jnkku-style-anima-krea2/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+jnkku-style-anima-krea2/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+heavy-makeup/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+heavy-makeup/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+heavy-makeup/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+heavy-makeup/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+heavy-makeup/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+heavy-makeup/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+dgirls-multiple-females/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+dgirls-multiple-females/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+brightness-slider-krea2/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+brightness-slider-krea2/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+brightness-slider-krea2/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+brightness-slider-krea2/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+brightness-slider-krea2/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+brightness-slider-krea2/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krealighteffect/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krealighteffect/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krealighteffect/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krealighteffect/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krealighteffect/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+krealighteffect/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+valiant-hearts-the-great-war-game-styles/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+valiant-hearts-the-great-war-game-styles/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+valiant-hearts-the-great-war-game-styles/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+valiant-hearts-the-great-war-game-styles/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+valiant-hearts-the-great-war-game-styles/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+valiant-hearts-the-great-war-game-styles/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text
+bone-armor-illustrious/example_images/01_image.jpeg filter=lfs diff=lfs merge=lfs -text
+bone-armor-illustrious/example_images/02_image.jpeg filter=lfs diff=lfs merge=lfs -text
+bone-armor-illustrious/example_images/03_image.jpeg filter=lfs diff=lfs merge=lfs -text
+bone-armor-illustrious/example_images/04_image.jpeg filter=lfs diff=lfs merge=lfs -text
+bone-armor-illustrious/example_images/05_image.jpeg filter=lfs diff=lfs merge=lfs -text
+bone-armor-illustrious/example_images/06_image.jpeg filter=lfs diff=lfs merge=lfs -text

bb-dubai-fate-series/bbdubai_krea2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff509f6198eb37eed444f23326fd61bf9ec53d944186a70281fc386448bdf735
+size 58815064

bb-dubai-fate-series/example_images/NO_PREVIEWS.txt

@@ -0,0 +1,9 @@
+No example media was saved for this LoRA.
+
+Every preview CivitAI holds for this version is rated above the
+'Soft' ceiling set by media.max_nsfw_level in config/collection.yaml.
+The gallery is uploaded by viewers, so this reflects what people chose
+to post, not what the LoRA itself produces.
+
+To fetch them anyway, raise media.max_nsfw_level (1=PG 2=PG-13 4=R
+8=X 16=XXX) and re-run:  python -m lorakit build

bb-dubai-fate-series/info.txt

@@ -0,0 +1,266 @@
+═══════════════════════════════════════════════════════════════════════════════
+  BB DUBAI - FATE SERIES
+═══════════════════════════════════════════════════════════════════════════════
+
+───────────────────────────────────────────────────────────────────────────────
+  IDENTITY
+───────────────────────────────────────────────────────────────────────────────
+  Base model        Krea 2
+  Generation mode   not declared by the author
+  Checkpoint        not determined — verify before use
+  Category          Character / subject
+  Author            extrememittens
+  Source            https://civitai.com/models/841358
+  Provenance        CivitAI · model 841358 · version 3286189
+
+───────────────────────────────────────────────────────────────────────────────
+  WEIGHTS
+───────────────────────────────────────────────────────────────────────────────
+  File              bbdubai_krea2.safetensors
+  Version           krea2
+  Size              56.1 MB
+  Format            SafeTensor / unknown precision
+  SHA256            ff509f6198eb37eed444f23326fd61bf9ec53d944186a70281fc386448bdf735
+  Published         2026-09-01
+
+───────────────────────────────────────────────────────────────────────────────
+  TRIGGER WORDS
+───────────────────────────────────────────────────────────────────────────────
+  › "bbdubai"
+  › "long, purple hair styled in braided pigtails, featuring a yellow
+    ribbon"
+
+───────────────────────────────────────────────────────────────────────────────
+  RECOMMENDED SETTINGS
+───────────────────────────────────────────────────────────────────────────────
+  The author did not publish specific settings.
+  No author settings were published. Community baseline: steps Turbo: 8
+  (official README and HF card CLI example; the ComfyUI template ships 8).
+  Raw: 52 (official README). The krea-2 CLI's generic default is 28.
+  Community Turbo LoRAs commonly suggest 8-12 (e.g. the Identity Edit card
+  recommends Turbo at 8-12, or Raw at 20 for removals). A community 4-step
+  LoRA on Turbo makes 4 valid; its author calls 2 steps an out-of-spec
+  preview mode., cfg Turbo: CFG disabled - guidance_scale 0.0 in diffusers
+  and --cfg 0.0 in the krea-2 CLI, which is cfg 1.0 in ComfyUI's KSampler
+  (the official template ships cfg 1). These are the same thing; do not set
+  ComfyUI cfg to 0. Raw: CFG 3.5 (official README; CLI generic default 4.5).
+  Negative prompting on Turbo is effectively unused - the official templates
+  route ConditioningZeroOut into the negative slot., sampler ComfyUI:
+  KSampler with sampler euler, scheduler simple, denoise 1.0 (verified in
+  image_krea2_turbo_t2i.json: [seed, 'randomize', 8, 1, 'euler', 'simple',
+  1]). Timestep shift mu = 1.15 for Turbo; ComfyUI bakes this in via
+  Krea2.sampling_settings = {"multiplier": 1.0, "shift": 1.15}, so the plain
+  t2i template needs no ModelSampling node, while the style-reference
+  template adds ModelSamplingFlux at max_shift 1.15 / base_shift 0.5. The CLI
+  exposes --mu 1.15 for Turbo; Raw uses a resolution-derived mu interpolated
+  between --y1 0.5 and --y2 1.15 (ai-toolkit mirrors this as base_shift 0.5 /
+  max_shift 1.15 over image_seq_len 256-6400, exponential time shift).,
+  resolution 1024x1024 default. Turbo: 1K-2K (the official CLI example
+  generates 2048x2048; ComfyUI's ResolutionSelector defaults to 1 megapixel,
+  set 2.0 for 2K). Raw: "trained to generate upto 1k resolution".
+  Width/height are padded up to a multiple of 16., lora_strength 1.0 for
+  Krea's nine official style LoRAs (docs.comfy.org lists Recommended Strength
+  1.0 for every one), though the shipped t2i template's LoraLoaderModelOnly
+  is set to 0.8. The style-reference LoRA is used at 1.0 in its template. The
+  4-step distill LoRA documents 1.0 as its trained point, 1.0-1.5 as coherent
+  extrapolation and above ~1.5 as speckle artefacting. Community subject
+  LoRAs are typically recommended around 0.8-1.2 and dropped toward 0.6-0.9
+  when several are stacked..
+
+───────────────────────────────────────────────────────────────────────────────
+  ADOPTION & POPULARITY
+───────────────────────────────────────────────────────────────────────────────
+  Downloads               923
+  Rating                  100% positive (69/69)
+  Comments                0
+  Published               4 days ago
+  Download rate           26.2/day since release
+
+  Adoption rank           #160 of 200 in this collection  (Niche)
+  Momentum rank           #160 of 200 by download rate  (Long tail · fresh release)
+
+  Adoption tier is the percentile of total downloads within this collection;
+  momentum is the percentile of downloads-per-day since release. Momentum is
+  ranked rather than measured against a fixed rate, because this base model
+  is itself new and every LoRA looks fast on an absolute scale. Both numbers
+  are CivitAI's, read on the scrape date at the foot of this file.
+
+───────────────────────────────────────────────────────────────────────────────
+  CHAINING / STACKING
+───────────────────────────────────────────────────────────────────────────────
+  Stacks with others      not stated
+
+  The author gave no chaining guidance. The rules below are the
+  general Krea 2 ones, not anything specific to this LoRA.
+
+  General Krea 2 rules that apply here (see REFERENCE.txt):
+
+  • The most important rule is not about chaining but about which
+    checkpoint you chain onto: the official README says in bold "TRAIN on
+    Raw and RUN on Turbo", and "We highly recommend using RAW for training
+    LoRAs and applying them on Turbo for inference." LoRAs trained on Raw
+    are stated to work well on Turbo.
+  • Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2
+    LoRAs are DiT-only, so there is no CLIP side to chain and the text-
+    encoder branch stays untouched however many LoRAs you add.
+  • Krea 2 LoRAs stack in practice and the community treats multi-LoRA as
+    routine - style + character, style + slider, realism + character.
+    Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one
+    writes "I recently stacked over 20 of my own LoRAs on a single
+    generation and the output still came out great."
+  • Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge
+    on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is
+    loaded. The CivitAI 12GB training guide also notes rank affects
+    composability: "rank 8 produced a lighter style influence that was
+    easier to stack with other LoRAs, rank 16 provided a useful balance
+    between concept strength and composability, and rank 32 was better
+    suited to a precise object or concept intended to take priority over
+    other stacked LoRAs."
+  • Accelerator LoRAs are position-sensitive in intent, not in node order:
+    a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a
+    sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-
+    Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is
+    explicitly recommended by the 4-step LoRA's author.
+  • Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI
+    (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card
+    says "Keep mu = 1.15. The training targets are anchored to that grid; a
+    different shift moves the [sampler off them]"; the same reasoning
+    applies to any Turbo-trained adapter in the chain.
+  • Reference/edit LoRAs are designed to be stacked under ordinary subject
+    and style LoRAs - the Identity Edit card states "Composes with your
+    LoRAs: character/body/style LoRAs stack on top and steer the prior" and
+    recommends a stacked subject LoRA when unusual subjects drift toward
+    the base prior.
+  • Krea 2 LoRAs cannot be chained with LoRAs for any other base model.
+    Many CivitAI pages carry multi-model titles like "(Krea-2 + ZIT)" or
+    "[Flux | ZIB | Krea2]" - those are separate files for separate
+    architectures shipped under one model page, not a combined adapter.
+  • To combine two full Krea 2 checkpoints rather than adapters, ComfyUI
+    ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-
+    section weights for first., tmlp., txtmlp., tproj.,
+    txtfusion.projector, txtfusion.layerwise_blocks.0-1,
+    txtfusion.refiner_blocks.0-1, blocks.0-27 and last.
+  • The author never states which weight family this targets. Silent
+    degradation, never a hard error, between Raw and Turbo. The two share
+    one architecture - ai-toolkit's config is commented "The reference
+    'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)"
+    - so every LoRA tensor shape matches on both and ComfyUI/diffuser
+
+───────────────────────────────────────────────────────────────────────────────
+  PROS
+───────────────────────────────────────────────────────────────────────────────
+  + Positively received — 69 thumbs-up with no down-votes. [source: CivitAI
+    stats]
+  + Explicit trigger word(s) — bbdubai, long, purple hair styled in braided
+    pigtails, featuring a yellow ribbon — so the effect can be turned on
+    and off from the prompt. [source: model metadata]
+  + Compact at 56 MB — cheap to stack with other LoRAs. [source: file size]
+
+───────────────────────────────────────────────────────────────────────────────
+  CONS & CAVEATS
+───────────────────────────────────────────────────────────────────────────────
+  - Author never states which generation mode it was trained for; test
+    against your own workflow before relying on it. [source: description]
+  - Thin documentation — expect trial and error. [source: description
+    length]
+  - Modest adoption (923 downloads) — less community feedback to rely on.
+    [source: source stats]
+
+───────────────────────────────────────────────────────────────────────────────
+  COMPATIBILITY WARNING
+───────────────────────────────────────────────────────────────────────────────
+  Silent degradation, never a hard error, between Raw and Turbo. The two
+  share one architecture - ai-toolkit's config is commented "The reference
+  'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)" - so
+  every LoRA tensor shape matches on both and ComfyUI/diffusers will load an
+  adapter against either without complaint. The lvladikov 4-step LoRA card
+  states this for the reverse direction: "Every layer it targets also exists
+  in Krea 2 Raw, so it will load there without complaint - but that is a side
+  effect of the shared architecture, not a supported mode", and reports
+  mixed, subject-dependent results with malformed subjects (duplicated heads,
+  fused limbs) at low step counts, needing 14+ steps with Raw's CFG on before
+  output is coherent. The intended direction (train on Raw, run on Turbo) is
+  officially blessed. The more common practical failure is a settings
+  mismatch rather than a weight mismatch: running a Turbo-targeted LoRA at
+  Raw's 52 steps + CFG 3.5, or leaving CFG enabled on Turbo, degrades output
+  with no warning. Loading a Krea 2 LoRA against a different model family
+  entirely (Flux, Qwen-Image, Z-Image) is also not an error - ComfyUI logs
+  "lora key not loaded: {key}" per key and generates as if no LoRA were
+  applied.
+
+───────────────────────────────────────────────────────────────────────────────
+  AUTHOR'S DESCRIPTION
+───────────────────────────────────────────────────────────────────────────────
+  A fate servant variation with some fun features..
+
+───────────────────────────────────────────────────────────────────────────────
+  EXAMPLE MEDIA
+───────────────────────────────────────────────────────────────────────────────
+  No work-safe previews were published for this version.
+
+───────────────────────────────────────────────────────────────────────────────
+  HOW TO USE (COMFYUI)
+───────────────────────────────────────────────────────────────────────────────
+  1. Update ComfyUI to at least v0.26.0 (released 2026-06-23), which added
+  local Krea2 support. For reference-image / style-reference / edit LoRAs you
+  need at least v0.29.0 (2026-07-29), which added regular and timestep-zero
+  reference images for Krea 2. ModelMergeKrea2 arrived in v0.27.0
+  (2026-06-30).
+
+  2. Put the base stack in place (files from huggingface.co/Comfy-Org/Krea-2,
+  which is not gated):
+     - ComfyUI/models/diffusion_models/krea2_turbo_fp8_scaled.safetensors
+  (documented recommended default; the repo also holds krea2_turbo_bf16,
+  _int8_convrot, _mxfp8, _nvfp4 and krea2_raw_bf16 / _fp8_scaled /
+  _int8_convrot)
+     - ComfyUI/models/text_encoders/qwen3vl_4b_fp8_scaled.safetensors (bf16
+  also available)
+     - ComfyUI/models/vae/qwen_image_vae.safetensors
+
+  3. Put the LoRA .safetensors in ComfyUI/models/loras/ - flat, no per-model
+  subfolder required.
+
+  4. Build or load the graph. The official image_krea2_turbo_t2i.json
+  template is: UNETLoader (krea2_turbo_fp8_scaled.safetensors) ->
+  LoraLoaderModelOnly -> KSampler; CLIPLoader
+  (qwen3vl_4b_fp8_scaled.safetensors, type "krea2") -> text encode ->
+  KSampler positive, with ConditioningZeroOut into negative; VAELoader
+  (qwen_image_vae.safetensors) -> VAEDecode -> SaveImage; EmptyLatentImage
+  1024x1024 driven by ResolutionSelector. An int8 sibling template
+  (image_krea2_turbo_t2i_int8.json) also ships.
+
+  5. Load the LoRA with LoraLoaderModelOnly, not LoraLoader - both official
+  templates do. Krea 2 LoRAs patch the DiT only; note that the DiT's own
+  text_fusion / txtfusion blocks are a normal LoRA target (the official style
+  LoRAs patch them), but the Qwen3-VL text encoder itself is not. Set
+  strength 1.0 for Krea's official style LoRAs (the shipped template uses
+  0.8) and add the LoRA's trigger word to the prompt.
+
+  6. Sampler settings: steps 8, cfg 1, sampler euler, scheduler simple,
+  denoise 1.0. Do not add a ModelSampling node for plain t2i - ComfyUI
+  already applies shift 1.15 from Krea2.sampling_settings.
+
+  7. For style reference, use the dedicated template: UNETLoader
+  krea2_turbo_int8_convrot.safetensors, LoraLoaderModelOnly
+  krea2_style_reference.safetensors at 1.0, ModelSamplingFlux (1.15, 0.5),
+  TextEncodeQwenImageEditPlus fed by LoadImage,
+  FluxKontextMultiReferenceLatentMethod set to index_timestep_zero, and
+  SamplerCustomAdvanced with KSamplerSelect euler + BasicScheduler simple/8 +
+  CFGGuider cfg 1.
+
+  8. For instruction editing (Krea 2 Identity Edit), stock nodes are not
+  enough: install the ComfyUI-Krea2Edit node pack, which ships its own
+  workflows, and run Turbo at 8-12 steps / cfg 1 (or Raw at 20 steps / cfg 3
+  for removals) with LoRA strength 1.0.
+
+  9. Low-VRAM alternative: install city96's ComfyUI-GGUF and use a community
+  GGUF build (realrebelai/KREA-2_GGUFs, quantized from Turbo;
+  molbal/krea2-gguf, quantized from Raw) with the Unet Loader (GGUF) node in
+  place of UNETLoader. There is no official GGUF release. LoRA loading is
+  unchanged.
+
+───────────────────────────────────────────────────────────────────────────────
+  Scraped by lorakit on 2026-09-05 04:00 UTC.
+  Stats and description are the author's; pros/cons are derived from the
+  evidence tagged beside each line.
+───────────────────────────────────────────────────────────────────────────────

bb-dubai-fate-series/metadata.json

@@ -0,0 +1,98 @@
+{
+  "name": "BB Dubai - Fate series",
+  "slug": "bb-dubai-fate-series",
+  "base_model": "Krea 2",
+  "media_type": "image",
+  "model_id": 841358,
+  "version_id": 3286189,
+  "version_name": "krea2",
+  "source": "civitai",
+  "url": "https://civitai.com/models/841358",
+  "repo_id": null,
+  "file_path": null,
+  "author": "extrememittens",
+  "category": {
+    "key": "character",
+    "label": "Character / subject"
+  },
+  "modalities": [],
+  "primary_modality": "unspecified",
+  "checkpoint_family": "unknown",
+  "trigger_words": [
+    "bbdubai",
+    "long, purple hair styled in braided pigtails, featuring a yellow ribbon"
+  ],
+  "recommended_settings": {},
+  "chaining": {
+    "author": {
+      "stacks": "not stated",
+      "notes": "",
+      "avoid": [],
+      "requires": [],
+      "order": "",
+      "strength": "",
+      "stated": false
+    },
+    "general_rules": [
+      "The most important rule is not about chaining but about which checkpoint you chain onto: the official README says in bold \"TRAIN on Raw and RUN on Turbo\", and \"We highly recommend using RAW for training LoRAs and applying them on Turbo for inference.\" LoRAs trained on Raw are stated to work well on Turbo.",
+      "Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2 LoRAs are DiT-only, so there is no CLIP side to chain and the text-encoder branch stays untouched however many LoRAs you add.",
+      "Krea 2 LoRAs stack in practice and the community treats multi-LoRA as routine - style + character, style + slider, realism + character. Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one writes \"I recently stacked over 20 of my own LoRAs on a single generation and the output still came out great.\"",
+      "Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is loaded. The CivitAI 12GB training guide also notes rank affects composability: \"rank 8 produced a lighter style influence that was easier to stack with other LoRAs, rank 16 provided a useful balance between concept strength and composability, and rank 32 was better suited to a precise object or concept intended to take priority over other stacked LoRAs.\"",
+      "Accelerator LoRAs are position-sensitive in intent, not in node order: a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is explicitly recommended by the 4-step LoRA's author.",
+      "Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card says \"Keep mu = 1.15. The training targets are anchored to that grid; a different shift moves the [sampler off them]\"; the same reasoning applies to any Turbo-trained adapter in the chain.",
+      "Reference/edit LoRAs are designed to be stacked under ordinary subject and style LoRAs - the Identity Edit card states \"Composes with your LoRAs: character/body/style LoRAs stack on top and steer the prior\" and recommends a stacked subject LoRA when unusual subjects drift toward the base prior.",
+      "Krea 2 LoRAs cannot be chained with LoRAs for any other base model. Many CivitAI pages carry multi-model titles like \"(Krea-2 + ZIT)\" or \"[Flux | ZIB | Krea2]\" - those are separate files for separate architectures shipped under one model page, not a combined adapter.",
+      "To combine two full Krea 2 checkpoints rather than adapters, ComfyUI ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-section weights for first., tmlp., txtmlp., tproj., txtfusion.projector, txtfusion.layerwise_blocks.0-1, txtfusion.refiner_blocks.0-1, blocks.0-27 and last.",
+      "The author never states which weight family this targets. Silent degradation, never a hard error, between Raw and Turbo. The two share one architecture - ai-toolkit's config is commented \"The reference 'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)\" - so every LoRA tensor shape matches on both and ComfyUI/diffuser"
+    ]
+  },
+  "pros": [
+    "Positively received — 69 thumbs-up with no down-votes. [source: CivitAI stats]",
+    "Explicit trigger word(s) — bbdubai, long, purple hair styled in braided pigtails, featuring a yellow ribbon — so the effect can be turned on and off from the prompt. [source: model metadata]",
+    "Compact at 56 MB — cheap to stack with other LoRAs. [source: file size]"
+  ],
+  "cons": [
+    "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]",
+    "Thin documentation — expect trial and error. [source: description length]",
+    "Modest adoption (923 downloads) — less community feedback to rely on. [source: source stats]"
+  ],
+  "file": {
+    "name": "bbdubai_krea2.safetensors",
+    "size_bytes": 58815064,
+    "size_mb": 56.09,
+    "sha256": "ff509f6198eb37eed444f23326fd61bf9ec53d944186a70281fc386448bdf735"
+  },
+  "stats": {
+    "downloads": 923,
+    "thumbs_up": 69,
+    "thumbs_down": 0
+  },
+  "popularity": {
+    "downloads": 923,
+    "thumbs_up": 69,
+    "thumbs_down": 0,
+    "comments": 0,
+    "approval": "100% positive (69/69)",
+    "age_days": 3.9,
+    "downloads_per_day": 26.25,
+    "rank_by_downloads": 160,
+    "rank_by_momentum": 160,
+    "of_total": 200,
+    "adoption_tier": "Niche",
+    "momentum": "Long tail · fresh release"
+  },
+  "selection_score": 84.7,
+  "selection_reasons": [
+    "923 downloads",
+    "69 thumbs-up",
+    "top-ranked for its category"
+  ],
+  "example_media": [],
+  "tags": [
+    "fate grand order",
+    "character",
+    "fate_go"
+  ],
+  "scraped_at": "2026-09-05T04:00:23+00:00",
+  "path": "krea-2/bb-dubai-fate-series"
+}

bone-armor-illustrious/Bone_Armor__Krea_2_V1__epoch_10.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90b04e8312b55de68bed5ec5aefff474389cbd63670e41eb092e53df0a9fc4c1
+size 114326424

bone-armor-illustrious/info.txt

@@ -0,0 +1,255 @@
+═══════════════════════════════════════════════════════════════════════════════
+  BONE ARMOR [ILLUSTRIOUS]
+═══════════════════════════════════════════════════════════════════════════════
+
+───────────────────────────────────────────────────────────────────────────────
+  IDENTITY
+───────────────────────────────────────────────────────────────────────────────
+  Base model        Krea 2
+  Generation mode   not declared by the author
+  Checkpoint        not determined — verify before use
+  Category          Visual style
+  Author            GrimGr1m
+  Source            https://civitai.com/models/2621916
+  Provenance        CivitAI · model 2621916 · version 3295635
+
+───────────────────────────────────────────────────────────────────────────────
+  WEIGHTS
+───────────────────────────────────────────────────────────────────────────────
+  File              Bone_Armor__Krea_2_V1__epoch_10.safetensors
+  Version           Krea 2 V1
+  Size              109.0 MB
+  Format            SafeTensor / unknown precision
+  SHA256            90b04e8312b55de68bed5ec5aefff474389cbd63670e41eb092e53df0a9fc4c1
+  Published         2026-09-04
+
+───────────────────────────────────────────────────────────────────────────────
+  TRIGGER WORDS
+───────────────────────────────────────────────────────────────────────────────
+  › "bone armor, skeleton armor,"
+  › "mask,"
+  › "helmet,"
+
+───────────────────────────────────────────────────────────────────────────────
+  RECOMMENDED SETTINGS
+───────────────────────────────────────────────────────────────────────────────
+  Taken verbatim from the author's description:
+
+  LoRA strength       0.8–1.0
+
+───────────────────────────────────────────────────────────────────────────────
+  ADOPTION & POPULARITY
+───────────────────────────────────────────────────────────────────────────────
+  Downloads               338
+  Rating                  100% positive (63/63)
+  Comments                0
+  Published               1 days ago
+  Download rate           38.0/day since release
+
+  Adoption rank           #200 of 200 in this collection  (Niche)
+  Momentum rank           #99 of 200 by download rate  (Steady · fresh release)
+
+  Adoption tier is the percentile of total downloads within this collection;
+  momentum is the percentile of downloads-per-day since release. Momentum is
+  ranked rather than measured against a fixed rate, because this base model
+  is itself new and every LoRA looks fast on an absolute scale. Both numbers
+  are CivitAI's, read on the scrape date at the foot of this file.
+
+───────────────────────────────────────────────────────────────────────────────
+  CHAINING / STACKING
+───────────────────────────────────────────────────────────────────────────────
+  Stacks with others      not stated
+
+  The author gave no chaining guidance. The rules below are the
+  general Krea 2 ones, not anything specific to this LoRA.
+
+  General Krea 2 rules that apply here (see REFERENCE.txt):
+
+  • The most important rule is not about chaining but about which
+    checkpoint you chain onto: the official README says in bold "TRAIN on
+    Raw and RUN on Turbo", and "We highly recommend using RAW for training
+    LoRAs and applying them on Turbo for inference." LoRAs trained on Raw
+    are stated to work well on Turbo.
+  • Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2
+    LoRAs are DiT-only, so there is no CLIP side to chain and the text-
+    encoder branch stays untouched however many LoRAs you add.
+  • Krea 2 LoRAs stack in practice and the community treats multi-LoRA as
+    routine - style + character, style + slider, realism + character.
+    Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one
+    writes "I recently stacked over 20 of my own LoRAs on a single
+    generation and the output still came out great."
+  • Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge
+    on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is
+    loaded. The CivitAI 12GB training guide also notes rank affects
+    composability: "rank 8 produced a lighter style influence that was
+    easier to stack with other LoRAs, rank 16 provided a useful balance
+    between concept strength and composability, and rank 32 was better
+    suited to a precise object or concept intended to take priority over
+    other stacked LoRAs."
+  • Accelerator LoRAs are position-sensitive in intent, not in node order:
+    a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a
+    sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-
+    Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is
+    explicitly recommended by the 4-step LoRA's author.
+  • Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI
+    (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card
+    says "Keep mu = 1.15. The training targets are anchored to that grid; a
+    different shift moves the [sampler off them]"; the same reasoning
+    applies to any Turbo-trained adapter in the chain.
+  • Reference/edit LoRAs are designed to be stacked under ordinary subject
+    and style LoRAs - the Identity Edit card states "Composes with your
+    LoRAs: character/body/style LoRAs stack on top and steer the prior" and
+    recommends a stacked subject LoRA when unusual subjects drift toward
+    the base prior.
+  • Krea 2 LoRAs cannot be chained with LoRAs for any other base model.
+    Many CivitAI pages carry multi-model titles like "(Krea-2 + ZIT)" or
+    "[Flux | ZIB | Krea2]" - those are separate files for separate
+    architectures shipped under one model page, not a combined adapter.
+  • To combine two full Krea 2 checkpoints rather than adapters, ComfyUI
+    ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-
+    section weights for first., tmlp., txtmlp., tproj.,
+    txtfusion.projector, txtfusion.layerwise_blocks.0-1,
+    txtfusion.refiner_blocks.0-1, blocks.0-27 and last.
+  • The author never states which weight family this targets. Silent
+    degradation, never a hard error, between Raw and Turbo. The two share
+    one architecture - ai-toolkit's config is commented "The reference
+    'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)"
+    - so every LoRA tensor shape matches on both and ComfyUI/diffuser
+
+───────────────────────────────────────────────────────────────────────────────
+  PROS
+───────────────────────────────────────────────────────────────────────────────
+  + Positively received — 63 thumbs-up with no down-votes. [source: CivitAI
+    stats]
+  + Explicit trigger word(s) — bone armor, skeleton armor,, mask,, helmet,
+    — so the effect can be turned on and off from the prompt. [source:
+    model metadata]
+  + Compact at 109 MB — cheap to stack with other LoRAs. [source: file
+    size]
+  + Author published concrete recommended settings. [source: description]
+
+───────────────────────────────────────────────────────────────────────────────
+  CONS & CAVEATS
+───────────────────────────────────────────────────────────────────────────────
+  - Author never states which generation mode it was trained for; test
+    against your own workflow before relying on it. [source: description]
+  - Modest adoption (338 downloads) — less community feedback to rely on.
+    [source: source stats]
+
+───────────────────────────────────────────────────────────────────────────────
+  COMPATIBILITY WARNING
+───────────────────────────────────────────────────────────────────────────────
+  Silent degradation, never a hard error, between Raw and Turbo. The two
+  share one architecture - ai-toolkit's config is commented "The reference
+  'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)" - so
+  every LoRA tensor shape matches on both and ComfyUI/diffusers will load an
+  adapter against either without complaint. The lvladikov 4-step LoRA card
+  states this for the reverse direction: "Every layer it targets also exists
+  in Krea 2 Raw, so it will load there without complaint - but that is a side
+  effect of the shared architecture, not a supported mode", and reports
+  mixed, subject-dependent results with malformed subjects (duplicated heads,
+  fused limbs) at low step counts, needing 14+ steps with Raw's CFG on before
+  output is coherent. The intended direction (train on Raw, run on Turbo) is
+  officially blessed. The more common practical failure is a settings
+  mismatch rather than a weight mismatch: running a Turbo-targeted LoRA at
+  Raw's 52 steps + CFG 3.5, or leaving CFG enabled on Turbo, degrades output
+  with no warning. Loading a Krea 2 LoRA against a different model family
+  entirely (Flux, Qwen-Image, Z-Image) is also not an error - ComfyUI logs
+  "lora key not loaded: {key}" per key and generates as if no LoRA were
+  applied.
+
+───────────────────────────────────────────────────────────────────────────────
+  AUTHOR'S DESCRIPTION
+───────────────────────────────────────────────────────────────────────────────
+  Prompts: bone armor, skeleton armor,
+
+  Optional Prompts: helmet, mask
+
+  Strength: 0.8-1.0 (change if necessary)
+
+  I love making these LoRA's for y'all. You can support me on Ko-fi . You can
+  also commission LoRAs and images from me!
+
+───────────────────────────────────────────────────────────────────────────────
+  EXAMPLE MEDIA
+───────────────────────────────────────────────────────────────────────────────
+  6 preview(s) in ./example_images/ — community results for
+  this LoRA. Videos keep a matching _poster.jpg still frame.
+
+  No prompts are recorded below: CivitAI's public API flags these posts as
+  having generation metadata but does not return it, so the prompts cannot be
+  scraped. Open the model page to read them.
+
+  01_image.jpeg  (image, 1024x1536)
+  02_image.jpeg  (image, 1024x1536)
+  03_image.jpeg  (image, 1024x1536)
+  04_image.jpeg  (image, 1024x1536)
+  05_image.jpeg  (image, 1024x1536)
+  06_image.jpeg  (image, 1024x1536)
+
+───────────────────────────────────────────────────────────────────────────────
+  HOW TO USE (COMFYUI)
+───────────────────────────────────────────────────────────────────────────────
+  1. Update ComfyUI to at least v0.26.0 (released 2026-06-23), which added
+  local Krea2 support. For reference-image / style-reference / edit LoRAs you
+  need at least v0.29.0 (2026-07-29), which added regular and timestep-zero
+  reference images for Krea 2. ModelMergeKrea2 arrived in v0.27.0
+  (2026-06-30).
+
+  2. Put the base stack in place (files from huggingface.co/Comfy-Org/Krea-2,
+  which is not gated):
+     - ComfyUI/models/diffusion_models/krea2_turbo_fp8_scaled.safetensors
+  (documented recommended default; the repo also holds krea2_turbo_bf16,
+  _int8_convrot, _mxfp8, _nvfp4 and krea2_raw_bf16 / _fp8_scaled /
+  _int8_convrot)
+     - ComfyUI/models/text_encoders/qwen3vl_4b_fp8_scaled.safetensors (bf16
+  also available)
+     - ComfyUI/models/vae/qwen_image_vae.safetensors
+
+  3. Put the LoRA .safetensors in ComfyUI/models/loras/ - flat, no per-model
+  subfolder required.
+
+  4. Build or load the graph. The official image_krea2_turbo_t2i.json
+  template is: UNETLoader (krea2_turbo_fp8_scaled.safetensors) ->
+  LoraLoaderModelOnly -> KSampler; CLIPLoader
+  (qwen3vl_4b_fp8_scaled.safetensors, type "krea2") -> text encode ->
+  KSampler positive, with ConditioningZeroOut into negative; VAELoader
+  (qwen_image_vae.safetensors) -> VAEDecode -> SaveImage; EmptyLatentImage
+  1024x1024 driven by ResolutionSelector. An int8 sibling template
+  (image_krea2_turbo_t2i_int8.json) also ships.
+
+  5. Load the LoRA with LoraLoaderModelOnly, not LoraLoader - both official
+  templates do. Krea 2 LoRAs patch the DiT only; note that the DiT's own
+  text_fusion / txtfusion blocks are a normal LoRA target (the official style
+  LoRAs patch them), but the Qwen3-VL text encoder itself is not. Set
+  strength 1.0 for Krea's official style LoRAs (the shipped template uses
+  0.8) and add the LoRA's trigger word to the prompt.
+
+  6. Sampler settings: steps 8, cfg 1, sampler euler, scheduler simple,
+  denoise 1.0. Do not add a ModelSampling node for plain t2i - ComfyUI
+  already applies shift 1.15 from Krea2.sampling_settings.
+
+  7. For style reference, use the dedicated template: UNETLoader
+  krea2_turbo_int8_convrot.safetensors, LoraLoaderModelOnly
+  krea2_style_reference.safetensors at 1.0, ModelSamplingFlux (1.15, 0.5),
+  TextEncodeQwenImageEditPlus fed by LoadImage,
+  FluxKontextMultiReferenceLatentMethod set to index_timestep_zero, and
+  SamplerCustomAdvanced with KSamplerSelect euler + BasicScheduler simple/8 +
+  CFGGuider cfg 1.
+
+  8. For instruction editing (Krea 2 Identity Edit), stock nodes are not
+  enough: install the ComfyUI-Krea2Edit node pack, which ships its own
+  workflows, and run Turbo at 8-12 steps / cfg 1 (or Raw at 20 steps / cfg 3
+  for removals) with LoRA strength 1.0.
+
+  9. Low-VRAM alternative: install city96's ComfyUI-GGUF and use a community
+  GGUF build (realrebelai/KREA-2_GGUFs, quantized from Turbo;
+  molbal/krea2-gguf, quantized from Raw) with the Unet Loader (GGUF) node in
+  place of UNETLoader. There is no official GGUF release. LoRA loading is
+  unchanged.
+
+───────────────────────────────────────────────────────────────────────────────
+  Scraped by lorakit on 2026-09-05 04:01 UTC.
+  Stats and description are the author's; pros/cons are derived from the
+  evidence tagged beside each line.
+───────────────────────────────────────────────────────────────────────────────

bone-armor-illustrious/metadata.json

@@ -0,0 +1,145 @@
+{
+  "name": "Bone Armor [Illustrious]",
+  "slug": "bone-armor-illustrious",
+  "base_model": "Krea 2",
+  "media_type": "image",
+  "model_id": 2621916,
+  "version_id": 3295635,
+  "version_name": "Krea 2 V1",
+  "source": "civitai",
+  "url": "https://civitai.com/models/2621916",
+  "repo_id": null,
+  "file_path": null,
+  "author": "GrimGr1m",
+  "category": {
+    "key": "style",
+    "label": "Visual style"
+  },
+  "modalities": [],
+  "primary_modality": "unspecified",
+  "checkpoint_family": "unknown",
+  "trigger_words": [
+    "bone armor, skeleton armor,",
+    "mask,",
+    "helmet,"
+  ],
+  "recommended_settings": {
+    "LoRA strength": "0.8–1.0"
+  },
+  "chaining": {
+    "author": {
+      "stacks": "not stated",
+      "notes": "",
+      "avoid": [],
+      "requires": [],
+      "order": "",
+      "strength": "",
+      "stated": false
+    },
+    "general_rules": [
+      "The most important rule is not about chaining but about which checkpoint you chain onto: the official README says in bold \"TRAIN on Raw and RUN on Turbo\", and \"We highly recommend using RAW for training LoRAs and applying them on Turbo for inference.\" LoRAs trained on Raw are stated to work well on Turbo.",
+      "Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2 LoRAs are DiT-only, so there is no CLIP side to chain and the text-encoder branch stays untouched however many LoRAs you add.",
+      "Krea 2 LoRAs stack in practice and the community treats multi-LoRA as routine - style + character, style + slider, realism + character. Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one writes \"I recently stacked over 20 of my own LoRAs on a single generation and the output still came out great.\"",
+      "Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is loaded. The CivitAI 12GB training guide also notes rank affects composability: \"rank 8 produced a lighter style influence that was easier to stack with other LoRAs, rank 16 provided a useful balance between concept strength and composability, and rank 32 was better suited to a precise object or concept intended to take priority over other stacked LoRAs.\"",
+      "Accelerator LoRAs are position-sensitive in intent, not in node order: a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is explicitly recommended by the 4-step LoRA's author.",
+      "Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card says \"Keep mu = 1.15. The training targets are anchored to that grid; a different shift moves the [sampler off them]\"; the same reasoning applies to any Turbo-trained adapter in the chain.",
+      "Reference/edit LoRAs are designed to be stacked under ordinary subject and style LoRAs - the Identity Edit card states \"Composes with your LoRAs: character/body/style LoRAs stack on top and steer the prior\" and recommends a stacked subject LoRA when unusual subjects drift toward the base prior.",
+      "Krea 2 LoRAs cannot be chained with LoRAs for any other base model. Many CivitAI pages carry multi-model titles like \"(Krea-2 + ZIT)\" or \"[Flux | ZIB | Krea2]\" - those are separate files for separate architectures shipped under one model page, not a combined adapter.",
+      "To combine two full Krea 2 checkpoints rather than adapters, ComfyUI ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-section weights for first., tmlp., txtmlp., tproj., txtfusion.projector, txtfusion.layerwise_blocks.0-1, txtfusion.refiner_blocks.0-1, blocks.0-27 and last.",
+      "The author never states which weight family this targets. Silent degradation, never a hard error, between Raw and Turbo. The two share one architecture - ai-toolkit's config is commented \"The reference 'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)\" - so every LoRA tensor shape matches on both and ComfyUI/diffuser"
+    ]
+  },
+  "pros": [
+    "Positively received — 63 thumbs-up with no down-votes. [source: CivitAI stats]",
+    "Explicit trigger word(s) — bone armor, skeleton armor,, mask,, helmet, — so the effect can be turned on and off from the prompt. [source: model metadata]",
+    "Compact at 109 MB — cheap to stack with other LoRAs. [source: file size]",
+    "Author published concrete recommended settings. [source: description]"
+  ],
+  "cons": [
+    "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]",
+    "Modest adoption (338 downloads) — less community feedback to rely on. [source: source stats]"
+  ],
+  "file": {
+    "name": "Bone_Armor__Krea_2_V1__epoch_10.safetensors",
+    "size_bytes": 114326424,
+    "size_mb": 109.03,
+    "sha256": "90b04e8312b55de68bed5ec5aefff474389cbd63670e41eb092e53df0a9fc4c1"
+  },
+  "stats": {
+    "downloads": 338,
+    "thumbs_up": 63,
+    "thumbs_down": 0
+  },
+  "popularity": {
+    "downloads": 338,
+    "thumbs_up": 63,
+    "thumbs_down": 0,
+    "comments": 0,
+    "approval": "100% positive (63/63)",
+    "age_days": 0.6,
+    "downloads_per_day": 38.0,
+    "rank_by_downloads": 200,
+    "rank_by_momentum": 99,
+    "of_total": 200,
+    "adoption_tier": "Niche",
+    "momentum": "Steady · fresh release"
+  },
+  "selection_score": 84.0,
+  "selection_reasons": [
+    "338 downloads",
+    "63 thumbs-up",
+    "top-ranked for its category"
+  ],
+  "example_media": [
+    {
+      "file": "01_image.jpeg",
+      "type": "image",
+      "resolution": "1024x1536",
+      "prompt": ""
+    },
+    {
+      "file": "02_image.jpeg",
+      "type": "image",
+      "resolution": "1024x1536",
+      "prompt": ""
+    },
+    {
+      "file": "03_image.jpeg",
+      "type": "image",
+      "resolution": "1024x1536",
+      "prompt": ""
+    },
+    {
+      "file": "04_image.jpeg",
+      "type": "image",
+      "resolution": "1024x1536",
+      "prompt": ""
+    },
+    {
+      "file": "05_image.jpeg",
+      "type": "image",
+      "resolution": "1024x1536",
+      "prompt": ""
+    },
+    {
+      "file": "06_image.jpeg",
+      "type": "image",
+      "resolution": "1024x1536",
+      "prompt": ""
+    }
+  ],
+  "tags": [
+    "anime",
+    "character",
+    "skeleton",
+    "bone",
+    "lora",
+    "characters",
+    "pokemon",
+    "armor",
+    "black",
+    "sexy"
+  ],
+  "scraped_at": "2026-09-05T04:01:21+00:00",
+  "path": "krea-2/bone-armor-illustrious"
+}

brightness-slider-krea2/brightness_krea2_loraholic.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87d7bfd546db8b040f39d50fbcc13c1cd6e2431a0ef2312c00673404a7bfab17
+size 7203656

brightness-slider-krea2/info.txt

@@ -0,0 +1,285 @@
+═══════════════════════════════════════════════════════════════════════════════
+  BRIGHTNESS SLIDER - KREA2
+═══════════════════════════════════════════════════════════════════════════════
+
+───────────────────────────────────────────────────────────────────────────────
+  IDENTITY
+───────────────────────────────────────────────────────────────────────────────
+  Base model        Krea 2
+  Generation mode   not declared by the author
+  Checkpoint        not determined — verify before use
+  Category          Uncategorised
+  Author            Loraholic
+  Source            https://civitai.com/models/2764196
+  Provenance        CivitAI · model 2764196 · version 3111088
+
+───────────────────────────────────────────────────────────────────────────────
+  WEIGHTS
+───────────────────────────────────────────────────────────────────────────────
+  File              brightness_krea2_loraholic.safetensors
+  Version           v1.0
+  Size              6.9 MB
+  Format            SafeTensor / unknown precision
+  SHA256            87d7bfd546db8b040f39d50fbcc13c1cd6e2431a0ef2312c00673404a7bfab17
+  Published         2026-07-08
+
+───────────────────────────────────────────────────────────────────────────────
+  TRIGGER WORDS
+───────────────────────────────────────────────────────────────────────────────
+  No trigger word. The LoRA applies as soon as it is loaded —
+  strength is the only way to dial the effect up or down.
+
+───────────────────────────────────────────────────────────────────────────────
+  RECOMMENDED SETTINGS
+───────────────────────────────────────────────────────────────────────────────
+  The author did not publish specific settings.
+  No author settings were published. Community baseline: steps Turbo: 8
+  (official README and HF card CLI example; the ComfyUI template ships 8).
+  Raw: 52 (official README). The krea-2 CLI's generic default is 28.
+  Community Turbo LoRAs commonly suggest 8-12 (e.g. the Identity Edit card
+  recommends Turbo at 8-12, or Raw at 20 for removals). A community 4-step
+  LoRA on Turbo makes 4 valid; its author calls 2 steps an out-of-spec
+  preview mode., cfg Turbo: CFG disabled - guidance_scale 0.0 in diffusers
+  and --cfg 0.0 in the krea-2 CLI, which is cfg 1.0 in ComfyUI's KSampler
+  (the official template ships cfg 1). These are the same thing; do not set
+  ComfyUI cfg to 0. Raw: CFG 3.5 (official README; CLI generic default 4.5).
+  Negative prompting on Turbo is effectively unused - the official templates
+  route ConditioningZeroOut into the negative slot., sampler ComfyUI:
+  KSampler with sampler euler, scheduler simple, denoise 1.0 (verified in
+  image_krea2_turbo_t2i.json: [seed, 'randomize', 8, 1, 'euler', 'simple',
+  1]). Timestep shift mu = 1.15 for Turbo; ComfyUI bakes this in via
+  Krea2.sampling_settings = {"multiplier": 1.0, "shift": 1.15}, so the plain
+  t2i template needs no ModelSampling node, while the style-reference
+  template adds ModelSamplingFlux at max_shift 1.15 / base_shift 0.5. The CLI
+  exposes --mu 1.15 for Turbo; Raw uses a resolution-derived mu interpolated
+  between --y1 0.5 and --y2 1.15 (ai-toolkit mirrors this as base_shift 0.5 /
+  max_shift 1.15 over image_seq_len 256-6400, exponential time shift).,
+  resolution 1024x1024 default. Turbo: 1K-2K (the official CLI example
+  generates 2048x2048; ComfyUI's ResolutionSelector defaults to 1 megapixel,
+  set 2.0 for 2K). Raw: "trained to generate upto 1k resolution".
+  Width/height are padded up to a multiple of 16., lora_strength 1.0 for
+  Krea's nine official style LoRAs (docs.comfy.org lists Recommended Strength
+  1.0 for every one), though the shipped t2i template's LoraLoaderModelOnly
+  is set to 0.8. The style-reference LoRA is used at 1.0 in its template. The
+  4-step distill LoRA documents 1.0 as its trained point, 1.0-1.5 as coherent
+  extrapolation and above ~1.5 as speckle artefacting. Community subject
+  LoRAs are typically recommended around 0.8-1.2 and dropped toward 0.6-0.9
+  when several are stacked..
+
+───────────────────────────────────────────────────────────────────────────────
+  ADOPTION & POPULARITY
+───────────────────────────────────────────────────────────────────────────────
+  Downloads               3,415
+  Rating                  100% positive (208/208)
+  Comments                7
+  Published               58 days ago
+  Download rate           58.5/day since release
+
+  Adoption rank           #80 of 200 in this collection  (Moderate)
+  Momentum rank           #68 of 200 by download rate  (Steady)
+
+  Adoption tier is the percentile of total downloads within this collection;
+  momentum is the percentile of downloads-per-day since release. Momentum is
+  ranked rather than measured against a fixed rate, because this base model
+  is itself new and every LoRA looks fast on an absolute scale. Both numbers
+  are CivitAI's, read on the scrape date at the foot of this file.
+
+───────────────────────────────────────────────────────────────────────────────
+  CHAINING / STACKING
+───────────────────────────────────────────────────────────────────────────────
+  Stacks with others      not stated
+
+  The author gave no chaining guidance. The rules below are the
+  general Krea 2 ones, not anything specific to this LoRA.
+
+  General Krea 2 rules that apply here (see REFERENCE.txt):
+
+  • The most important rule is not about chaining but about which
+    checkpoint you chain onto: the official README says in bold "TRAIN on
+    Raw and RUN on Turbo", and "We highly recommend using RAW for training
+    LoRAs and applying them on Turbo for inference." LoRAs trained on Raw
+    are stated to work well on Turbo.
+  • Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2
+    LoRAs are DiT-only, so there is no CLIP side to chain and the text-
+    encoder branch stays untouched however many LoRAs you add.
+  • Krea 2 LoRAs stack in practice and the community treats multi-LoRA as
+    routine - style + character, style + slider, realism + character.
+    Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one
+    writes "I recently stacked over 20 of my own LoRAs on a single
+    generation and the output still came out great."
+  • Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge
+    on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is
+    loaded. The CivitAI 12GB training guide also notes rank affects
+    composability: "rank 8 produced a lighter style influence that was
+    easier to stack with other LoRAs, rank 16 provided a useful balance
+    between concept strength and composability, and rank 32 was better
+    suited to a precise object or concept intended to take priority over
+    other stacked LoRAs."
+  • Accelerator LoRAs are position-sensitive in intent, not in node order:
+    a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a
+    sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-
+    Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is
+    explicitly recommended by the 4-step LoRA's author.
+  • Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI
+    (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card
+    says "Keep mu = 1.15. The training targets are anchored to that grid; a
+    different shift moves the [sampler off them]"; the same reasoning
+    applies to any Turbo-trained adapter in the chain.
+  • Reference/edit LoRAs are designed to be stacked under ordinary subject
+    and style LoRAs - the Identity Edit card states "Composes with your
+    LoRAs: character/body/style LoRAs stack on top and steer the prior" and
+    recommends a stacked subject LoRA when unusual subjects drift toward
+    the base prior.
+  • Krea 2 LoRAs cannot be chained with LoRAs for any other base model.
+    Many CivitAI pages carry multi-model titles like "(Krea-2 + ZIT)" or
+    "[Flux | ZIB | Krea2]" - those are separate files for separate
+    architectures shipped under one model page, not a combined adapter.
+  • To combine two full Krea 2 checkpoints rather than adapters, ComfyUI
+    ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-
+    section weights for first., tmlp., txtmlp., tproj.,
+    txtfusion.projector, txtfusion.layerwise_blocks.0-1,
+    txtfusion.refiner_blocks.0-1, blocks.0-27 and last.
+  • The author never states which weight family this targets. Silent
+    degradation, never a hard error, between Raw and Turbo. The two share
+    one architecture - ai-toolkit's config is commented "The reference
+    'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)"
+    - so every LoRA tensor shape matches on both and ComfyUI/diffuser
+
+───────────────────────────────────────────────────────────────────────────────
+  PROS
+───────────────────────────────────────────────────────────────────────────────
+  + Well established — 3,415 downloads. [source: CivitAI stats]
+  + Positively received — 208 thumbs-up with no down-votes. [source:
+    CivitAI stats]
+  + Compact at 7 MB — cheap to stack with other LoRAs. [source: file size]
+
+───────────────────────────────────────────────────────────────────────────────
+  CONS & CAVEATS
+───────────────────────────────────────────────────────────────────────────────
+  - No trigger word — the LoRA is always on once loaded, so strength is the
+    only control. [source: model metadata]
+  - Author never states which generation mode it was trained for; test
+    against your own workflow before relying on it. [source: description]
+
+───────────────────────────────────────────────────────────────────────────────
+  COMPATIBILITY WARNING
+───────────────────────────────────────────────────────────────────────────────
+  Silent degradation, never a hard error, between Raw and Turbo. The two
+  share one architecture - ai-toolkit's config is commented "The reference
+  'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)" - so
+  every LoRA tensor shape matches on both and ComfyUI/diffusers will load an
+  adapter against either without complaint. The lvladikov 4-step LoRA card
+  states this for the reverse direction: "Every layer it targets also exists
+  in Krea 2 Raw, so it will load there without complaint - but that is a side
+  effect of the shared architecture, not a supported mode", and reports
+  mixed, subject-dependent results with malformed subjects (duplicated heads,
+  fused limbs) at low step counts, needing 14+ steps with Raw's CFG on before
+  output is coherent. The intended direction (train on Raw, run on Turbo) is
+  officially blessed. The more common practical failure is a settings
+  mismatch rather than a weight mismatch: running a Turbo-targeted LoRA at
+  Raw's 52 steps + CFG 3.5, or leaving CFG enabled on Turbo, degrades output
+  with no warning. Loading a Krea 2 LoRA against a different model family
+  entirely (Flux, Qwen-Image, Z-Image) is also not an error - ComfyUI logs
+  "lora key not loaded: {key}" per key and generates as if no LoRA were
+  applied.
+
+───────────────────────────────────────────────────────────────────────────────
+  AUTHOR'S DESCRIPTION
+───────────────────────────────────────────────────────────────────────────────
+  Brightness Slider This is not the same as my Light Slider . This is more
+  like changing exposure, and the whole image change, instead of bumping
+  contrast like my light slider does.
+  Its a bit backward, in the sense that brighter is negative, but i suspect
+  most people going to use ut to make the image darker, so you can use pos
+  value. It is also the same direction as the Light Slider to make it
+  somewhat consistant accross LoRAs.
+  I will say the same as on my light slider: light LoRAs will almost always
+  degrade quality , at least a little. So its best to use prompting as the
+  main light controller, and use loras to push a little extra. Or use
+  dedicated brightness and contrast nodes instead of a LoRA.
+
+  Recommended range: -5.0 to 5.0
+
+───────────────────────────────────────────────────────────────────────────────
+  EXAMPLE MEDIA
+───────────────────────────────────────────────────────────────────────────────
+  6 preview(s) in ./example_images/ — community results for
+  this LoRA. Videos keep a matching _poster.jpg still frame.
+
+  No prompts are recorded below: CivitAI's public API flags these posts as
+  having generation metadata but does not return it, so the prompts cannot be
+  scraped. Open the model page to read them.
+
+  01_image.jpeg  (image, 2048x2560)
+  02_image.jpeg  (image, 2048x2560)
+  03_image.jpeg  (image, 1248x1824)
+  04_image.jpeg  (image, 2048x2560)
+  05_image.jpeg  (image, 1248x1824)
+  06_image.jpeg  (image, 2880x1728)
+
+───────────────────────────────────────────────────────────────────────────────
+  HOW TO USE (COMFYUI)
+───────────────────────────────────────────────────────────────────────────────
+  1. Update ComfyUI to at least v0.26.0 (released 2026-06-23), which added
+  local Krea2 support. For reference-image / style-reference / edit LoRAs you
+  need at least v0.29.0 (2026-07-29), which added regular and timestep-zero
+  reference images for Krea 2. ModelMergeKrea2 arrived in v0.27.0
+  (2026-06-30).
+
+  2. Put the base stack in place (files from huggingface.co/Comfy-Org/Krea-2,
+  which is not gated):
+     - ComfyUI/models/diffusion_models/krea2_turbo_fp8_scaled.safetensors
+  (documented recommended default; the repo also holds krea2_turbo_bf16,
+  _int8_convrot, _mxfp8, _nvfp4 and krea2_raw_bf16 / _fp8_scaled /
+  _int8_convrot)
+     - ComfyUI/models/text_encoders/qwen3vl_4b_fp8_scaled.safetensors (bf16
+  also available)
+     - ComfyUI/models/vae/qwen_image_vae.safetensors
+
+  3. Put the LoRA .safetensors in ComfyUI/models/loras/ - flat, no per-model
+  subfolder required.
+
+  4. Build or load the graph. The official image_krea2_turbo_t2i.json
+  template is: UNETLoader (krea2_turbo_fp8_scaled.safetensors) ->
+  LoraLoaderModelOnly -> KSampler; CLIPLoader
+  (qwen3vl_4b_fp8_scaled.safetensors, type "krea2") -> text encode ->
+  KSampler positive, with ConditioningZeroOut into negative; VAELoader
+  (qwen_image_vae.safetensors) -> VAEDecode -> SaveImage; EmptyLatentImage
+  1024x1024 driven by ResolutionSelector. An int8 sibling template
+  (image_krea2_turbo_t2i_int8.json) also ships.
+
+  5. Load the LoRA with LoraLoaderModelOnly, not LoraLoader - both official
+  templates do. Krea 2 LoRAs patch the DiT only; note that the DiT's own
+  text_fusion / txtfusion blocks are a normal LoRA target (the official style
+  LoRAs patch them), but the Qwen3-VL text encoder itself is not. Set
+  strength 1.0 for Krea's official style LoRAs (the shipped template uses
+  0.8) and add the LoRA's trigger word to the prompt.
+
+  6. Sampler settings: steps 8, cfg 1, sampler euler, scheduler simple,
+  denoise 1.0. Do not add a ModelSampling node for plain t2i - ComfyUI
+  already applies shift 1.15 from Krea2.sampling_settings.
+
+  7. For style reference, use the dedicated template: UNETLoader
+  krea2_turbo_int8_convrot.safetensors, LoraLoaderModelOnly
+  krea2_style_reference.safetensors at 1.0, ModelSamplingFlux (1.15, 0.5),
+  TextEncodeQwenImageEditPlus fed by LoadImage,
+  FluxKontextMultiReferenceLatentMethod set to index_timestep_zero, and
+  SamplerCustomAdvanced with KSamplerSelect euler + BasicScheduler simple/8 +
+  CFGGuider cfg 1.
+
+  8. For instruction editing (Krea 2 Identity Edit), stock nodes are not
+  enough: install the ComfyUI-Krea2Edit node pack, which ships its own
+  workflows, and run Turbo at 8-12 steps / cfg 1 (or Raw at 20 steps / cfg 3
+  for removals) with LoRA strength 1.0.
+
+  9. Low-VRAM alternative: install city96's ComfyUI-GGUF and use a community
+  GGUF build (realrebelai/KREA-2_GGUFs, quantized from Turbo;
+  molbal/krea2-gguf, quantized from Raw) with the Unet Loader (GGUF) node in
+  place of UNETLoader. There is no official GGUF release. LoRA loading is
+  unchanged.
+
+──────────────────────────────────────────────────────────────────────���────────
+  Scraped by lorakit on 2026-09-05 04:01 UTC.
+  Stats and description are the author's; pros/cons are derived from the
+  evidence tagged beside each line.
+───────────────────────────────────────────────────────────────────────────────

brightness-slider-krea2/metadata.json

@@ -0,0 +1,133 @@
+{
+  "name": "Brightness Slider - Krea2",
+  "slug": "brightness-slider-krea2",
+  "base_model": "Krea 2",
+  "media_type": "image",
+  "model_id": 2764196,
+  "version_id": 3111088,
+  "version_name": "v1.0",
+  "source": "civitai",
+  "url": "https://civitai.com/models/2764196",
+  "repo_id": null,
+  "file_path": null,
+  "author": "Loraholic",
+  "category": {
+    "key": "other",
+    "label": "Uncategorised"
+  },
+  "modalities": [],
+  "primary_modality": "unspecified",
+  "checkpoint_family": "unknown",
+  "trigger_words": [],
+  "recommended_settings": {},
+  "chaining": {
+    "author": {
+      "stacks": "not stated",
+      "notes": "",
+      "avoid": [],
+      "requires": [],
+      "order": "",
+      "strength": "",
+      "stated": false
+    },
+    "general_rules": [
+      "The most important rule is not about chaining but about which checkpoint you chain onto: the official README says in bold \"TRAIN on Raw and RUN on Turbo\", and \"We highly recommend using RAW for training LoRAs and applying them on Turbo for inference.\" LoRAs trained on Raw are stated to work well on Turbo.",
+      "Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2 LoRAs are DiT-only, so there is no CLIP side to chain and the text-encoder branch stays untouched however many LoRAs you add.",
+      "Krea 2 LoRAs stack in practice and the community treats multi-LoRA as routine - style + character, style + slider, realism + character. Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one writes \"I recently stacked over 20 of my own LoRAs on a single generation and the output still came out great.\"",
+      "Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is loaded. The CivitAI 12GB training guide also notes rank affects composability: \"rank 8 produced a lighter style influence that was easier to stack with other LoRAs, rank 16 provided a useful balance between concept strength and composability, and rank 32 was better suited to a precise object or concept intended to take priority over other stacked LoRAs.\"",
+      "Accelerator LoRAs are position-sensitive in intent, not in node order: a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is explicitly recommended by the 4-step LoRA's author.",
+      "Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card says \"Keep mu = 1.15. The training targets are anchored to that grid; a different shift moves the [sampler off them]\"; the same reasoning applies to any Turbo-trained adapter in the chain.",
+      "Reference/edit LoRAs are designed to be stacked under ordinary subject and style LoRAs - the Identity Edit card states \"Composes with your LoRAs: character/body/style LoRAs stack on top and steer the prior\" and recommends a stacked subject LoRA when unusual subjects drift toward the base prior.",
+      "Krea 2 LoRAs cannot be chained with LoRAs for any other base model. Many CivitAI pages carry multi-model titles like \"(Krea-2 + ZIT)\" or \"[Flux | ZIB | Krea2]\" - those are separate files for separate architectures shipped under one model page, not a combined adapter.",
+      "To combine two full Krea 2 checkpoints rather than adapters, ComfyUI ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-section weights for first., tmlp., txtmlp., tproj., txtfusion.projector, txtfusion.layerwise_blocks.0-1, txtfusion.refiner_blocks.0-1, blocks.0-27 and last.",
+      "The author never states which weight family this targets. Silent degradation, never a hard error, between Raw and Turbo. The two share one architecture - ai-toolkit's config is commented \"The reference 'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)\" - so every LoRA tensor shape matches on both and ComfyUI/diffuser"
+    ]
+  },
+  "pros": [
+    "Well established — 3,415 downloads. [source: CivitAI stats]",
+    "Positively received — 208 thumbs-up with no down-votes. [source: CivitAI stats]",
+    "Compact at 7 MB — cheap to stack with other LoRAs. [source: file size]"
+  ],
+  "cons": [
+    "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]",
+    "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]"
+  ],
+  "file": {
+    "name": "brightness_krea2_loraholic.safetensors",
+    "size_bytes": 7203656,
+    "size_mb": 6.87,
+    "sha256": "87d7bfd546db8b040f39d50fbcc13c1cd6e2431a0ef2312c00673404a7bfab17"
+  },
+  "stats": {
+    "downloads": 3415,
+    "thumbs_up": 208,
+    "thumbs_down": 0
+  },
+  "popularity": {
+    "downloads": 3415,
+    "thumbs_up": 208,
+    "thumbs_down": 0,
+    "comments": 7,
+    "approval": "100% positive (208/208)",
+    "age_days": 58.4,
+    "downloads_per_day": 58.51,
+    "rank_by_downloads": 80,
+    "rank_by_momentum": 68,
+    "of_total": 200,
+    "adoption_tier": "Moderate",
+    "momentum": "Steady"
+  },
+  "selection_score": 84.1,
+  "selection_reasons": [
+    "3,415 downloads",
+    "208 thumbs-up",
+    "top-ranked for its category"
+  ],
+  "example_media": [
+    {
+      "file": "01_image.jpeg",
+      "type": "image",
+      "resolution": "2048x2560",
+      "prompt": ""
+    },
+    {
+      "file": "02_image.jpeg",
+      "type": "image",
+      "resolution": "2048x2560",
+      "prompt": ""
+    },
+    {
+      "file": "03_image.jpeg",
+      "type": "image",
+      "resolution": "1248x1824",
+      "prompt": ""
+    },
+    {
+      "file": "04_image.jpeg",
+      "type": "image",
+      "resolution": "2048x2560",
+      "prompt": ""
+    },
+    {
+      "file": "05_image.jpeg",
+      "type": "image",
+      "resolution": "1248x1824",
+      "prompt": ""
+    },
+    {
+      "file": "06_image.jpeg",
+      "type": "image",
+      "resolution": "2880x1728",
+      "prompt": ""
+    }
+  ],
+  "tags": [
+    "loraholic",
+    "tool",
+    "shadow",
+    "light",
+    "slider"
+  ],
+  "scraped_at": "2026-09-05T04:01:14+00:00",
+  "path": "krea-2/brightness-slider-krea2"
+}

cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/Cheerleader_outfit_-_Krea_2_epoch_10.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f9a6764cb959fc46e10b2f142cffcbe92dfa8b5fe20b5c0abad71d98fc30980
+size 228588912

cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/example_images/06_video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9b074a20110b28474f554a2052d77cc866f28a6b1d75fa79db4b02e87e6f6b1
+size 1596857

cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/info.txt

@@ -0,0 +1,250 @@
+═══════════════════════════════════════════════════════════════════════════════
+  CHEERLEADER OUTFIT - (KREA 2, ANIMA, ILLUSTRIOUS, FLUX 2 KLEIN 9B, FLUX 2
+  KLEIN 4B & ZIT)
+═══════════════════════════════════════════════════════════════════════════════
+
+───────────────────────────────────────────────────────────────────────────────
+  IDENTITY
+───────────────────────────────────────────────────────────────────────────────
+  Base model        Krea 2
+  Generation mode   not declared by the author
+  Checkpoint        not determined — verify before use
+  Category          Clothing / wardrobe
+  Author            DTAL835217
+  Source            https://civitai.com/models/2610117
+  Provenance        CivitAI · model 2610117 · version 3200034
+
+───────────────────────────────────────────────────────────────────────────────
+  WEIGHTS
+───────────────────────────────────────────────────────────────────────────────
+  File              Cheerleader_outfit_-_Krea_2_epoch_10.safetensors
+  Version           Krea 2 Version
+  Size              218.0 MB
+  Format            SafeTensor / unknown precision
+  SHA256            6f9a6764cb959fc46e10b2f142cffcbe92dfa8b5fe20b5c0abad71d98fc30980
+  Published         2026-08-05
+
+───────────────────────────────────────────────────────────────────────────────
+  TRIGGER WORDS
+───────────────────────────────────────────────────────────────────────────────
+  No trigger word. The LoRA applies as soon as it is loaded —
+  strength is the only way to dial the effect up or down.
+
+───────────────────────────────────────────────────────────────────────────────
+  RECOMMENDED SETTINGS
+───────────────────────────────────────────────────────────────────────────────
+  Taken verbatim from the author's description:
+
+  LoRA strength       0.7–1.0
+
+───────────────────────────────────────────────────────────────────────────────
+  ADOPTION & POPULARITY
+───────────────────────────────────────────────────────────────────────────────
+  Downloads               2,602
+  Rating                  100% positive (127/127)
+  Comments                0
+  Published               31 days ago
+  Download rate           27.8/day since release
+
+  Adoption rank           #94 of 200 in this collection  (Moderate)
+  Momentum rank           #153 of 200 by download rate  (Long tail · fresh release)
+
+  Adoption tier is the percentile of total downloads within this collection;
+  momentum is the percentile of downloads-per-day since release. Momentum is
+  ranked rather than measured against a fixed rate, because this base model
+  is itself new and every LoRA looks fast on an absolute scale. Both numbers
+  are CivitAI's, read on the scrape date at the foot of this file.
+
+───────────────────────────────────────────────────────────────────────────────
+  CHAINING / STACKING
+───────────────────────────────────────────────────────────────────────────────
+  Stacks with others      not stated
+
+  The author gave no chaining guidance. The rules below are the
+  general Krea 2 ones, not anything specific to this LoRA.
+
+  General Krea 2 rules that apply here (see REFERENCE.txt):
+
+  • The most important rule is not about chaining but about which
+    checkpoint you chain onto: the official README says in bold "TRAIN on
+    Raw and RUN on Turbo", and "We highly recommend using RAW for training
+    LoRAs and applying them on Turbo for inference." LoRAs trained on Raw
+    are stated to work well on Turbo.
+  • Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2
+    LoRAs are DiT-only, so there is no CLIP side to chain and the text-
+    encoder branch stays untouched however many LoRAs you add.
+  • Krea 2 LoRAs stack in practice and the community treats multi-LoRA as
+    routine - style + character, style + slider, realism + character.
+    Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one
+    writes "I recently stacked over 20 of my own LoRAs on a single
+    generation and the output still came out great."
+  • Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge
+    on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is
+    loaded. The CivitAI 12GB training guide also notes rank affects
+    composability: "rank 8 produced a lighter style influence that was
+    easier to stack with other LoRAs, rank 16 provided a useful balance
+    between concept strength and composability, and rank 32 was better
+    suited to a precise object or concept intended to take priority over
+    other stacked LoRAs."
+  • Accelerator LoRAs are position-sensitive in intent, not in node order:
+    a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a
+    sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-
+    Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is
+    explicitly recommended by the 4-step LoRA's author.
+  • Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI
+    (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card
+    says "Keep mu = 1.15. The training targets are anchored to that grid; a
+    different shift moves the [sampler off them]"; the same reasoning
+    applies to any Turbo-trained adapter in the chain.
+  • Reference/edit LoRAs are designed to be stacked under ordinary subject
+    and style LoRAs - the Identity Edit card states "Composes with your
+    LoRAs: character/body/style LoRAs stack on top and steer the prior" and
+    recommends a stacked subject LoRA when unusual subjects drift toward
+    the base prior.
+  • Krea 2 LoRAs cannot be chained with LoRAs for any other base model.
+    Many CivitAI pages carry multi-model titles like "(Krea-2 + ZIT)" or
+    "[Flux | ZIB | Krea2]" - those are separate files for separate
+    architectures shipped under one model page, not a combined adapter.
+  • To combine two full Krea 2 checkpoints rather than adapters, ComfyUI
+    ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-
+    section weights for first., tmlp., txtmlp., tproj.,
+    txtfusion.projector, txtfusion.layerwise_blocks.0-1,
+    txtfusion.refiner_blocks.0-1, blocks.0-27 and last.
+  • The author never states which weight family this targets. Silent
+    degradation, never a hard error, between Raw and Turbo. The two share
+    one architecture - ai-toolkit's config is commented "The reference
+    'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)"
+    - so every LoRA tensor shape matches on both and ComfyUI/diffuser
+
+───────────────────────────────────────────────────────────────────────────────
+  PROS
+───────────────────────────────────────────────────────────────────────────────
+  + Well established — 2,602 downloads. [source: CivitAI stats]
+  + Positively received — 127 thumbs-up with no down-votes. [source:
+    CivitAI stats]
+  + Compact at 218 MB — cheap to stack with other LoRAs. [source: file
+    size]
+  + Author published concrete recommended settings. [source: description]
+
+───────────────────────────────────────────────────────────────────────────────
+  CONS & CAVEATS
+───────────────────────────────────────────────────────────────────────────────
+  - No trigger word — the LoRA is always on once loaded, so strength is the
+    only control. [source: model metadata]
+  - Author never states which generation mode it was trained for; test
+    against your own workflow before relying on it. [source: description]
+  - Thin documentation — expect trial and error. [source: description
+    length]
+
+───────────────────────────────────────────────────────────────────────────────
+  COMPATIBILITY WARNING
+───────────────────────────────────────────────────────────────────────────────
+  Silent degradation, never a hard error, between Raw and Turbo. The two
+  share one architecture - ai-toolkit's config is commented "The reference
+  'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)" - so
+  every LoRA tensor shape matches on both and ComfyUI/diffusers will load an
+  adapter against either without complaint. The lvladikov 4-step LoRA card
+  states this for the reverse direction: "Every layer it targets also exists
+  in Krea 2 Raw, so it will load there without complaint - but that is a side
+  effect of the shared architecture, not a supported mode", and reports
+  mixed, subject-dependent results with malformed subjects (duplicated heads,
+  fused limbs) at low step counts, needing 14+ steps with Raw's CFG on before
+  output is coherent. The intended direction (train on Raw, run on Turbo) is
+  officially blessed. The more common practical failure is a settings
+  mismatch rather than a weight mismatch: running a Turbo-targeted LoRA at
+  Raw's 52 steps + CFG 3.5, or leaving CFG enabled on Turbo, degrades output
+  with no warning. Loading a Krea 2 LoRA against a different model family
+  entirely (Flux, Qwen-Image, Z-Image) is also not an error - ComfyUI logs
+  "lora key not loaded: {key}" per key and generates as if no LoRA were
+  applied.
+
+───────────────────────────────────────────────────────────────────────────────
+  AUTHOR'S DESCRIPTION
+───────────────────────────────────────────────────────────────────────────────
+  A outfit lora based on replicating a cheerleader's outfit.
+  Recommended strength: 0.7 - 1.0.
+
+───────────────────────────────────────────────────────────────────────────────
+  EXAMPLE MEDIA
+───────────────────────────────────────────────────────────────────────────────
+  6 preview(s) in ./example_images/ — community results for
+  this LoRA. Videos keep a matching _poster.jpg still frame.
+
+  No prompts are recorded below: CivitAI's public API flags these posts as
+  having generation metadata but does not return it, so the prompts cannot be
+  scraped. Open the model page to read them.
+
+  01_image.jpeg  (image, 1024x768)
+  02_image.jpeg  (image, 832x1216)
+  03_image.jpeg  (image, 1024x768)
+  04_image.jpeg  (image, 832x1216)
+  05_image.jpeg  (image, 1024x768)
+  06_video.mp4  (video, 720x1280)
+      still: 06_poster.jpg
+
+───────────────────────────────────────────────────────────────────────────────
+  HOW TO USE (COMFYUI)
+───────────────────────────────────────────────────────────────────────────────
+  1. Update ComfyUI to at least v0.26.0 (released 2026-06-23), which added
+  local Krea2 support. For reference-image / style-reference / edit LoRAs you
+  need at least v0.29.0 (2026-07-29), which added regular and timestep-zero
+  reference images for Krea 2. ModelMergeKrea2 arrived in v0.27.0
+  (2026-06-30).
+
+  2. Put the base stack in place (files from huggingface.co/Comfy-Org/Krea-2,
+  which is not gated):
+     - ComfyUI/models/diffusion_models/krea2_turbo_fp8_scaled.safetensors
+  (documented recommended default; the repo also holds krea2_turbo_bf16,
+  _int8_convrot, _mxfp8, _nvfp4 and krea2_raw_bf16 / _fp8_scaled /
+  _int8_convrot)
+     - ComfyUI/models/text_encoders/qwen3vl_4b_fp8_scaled.safetensors (bf16
+  also available)
+     - ComfyUI/models/vae/qwen_image_vae.safetensors
+
+  3. Put the LoRA .safetensors in ComfyUI/models/loras/ - flat, no per-model
+  subfolder required.
+
+  4. Build or load the graph. The official image_krea2_turbo_t2i.json
+  template is: UNETLoader (krea2_turbo_fp8_scaled.safetensors) ->
+  LoraLoaderModelOnly -> KSampler; CLIPLoader
+  (qwen3vl_4b_fp8_scaled.safetensors, type "krea2") -> text encode ->
+  KSampler positive, with ConditioningZeroOut into negative; VAELoader
+  (qwen_image_vae.safetensors) -> VAEDecode -> SaveImage; EmptyLatentImage
+  1024x1024 driven by ResolutionSelector. An int8 sibling template
+  (image_krea2_turbo_t2i_int8.json) also ships.
+
+  5. Load the LoRA with LoraLoaderModelOnly, not LoraLoader - both official
+  templates do. Krea 2 LoRAs patch the DiT only; note that the DiT's own
+  text_fusion / txtfusion blocks are a normal LoRA target (the official style
+  LoRAs patch them), but the Qwen3-VL text encoder itself is not. Set
+  strength 1.0 for Krea's official style LoRAs (the shipped template uses
+  0.8) and add the LoRA's trigger word to the prompt.
+
+  6. Sampler settings: steps 8, cfg 1, sampler euler, scheduler simple,
+  denoise 1.0. Do not add a ModelSampling node for plain t2i - ComfyUI
+  already applies shift 1.15 from Krea2.sampling_settings.
+
+  7. For style reference, use the dedicated template: UNETLoader
+  krea2_turbo_int8_convrot.safetensors, LoraLoaderModelOnly
+  krea2_style_reference.safetensors at 1.0, ModelSamplingFlux (1.15, 0.5),
+  TextEncodeQwenImageEditPlus fed by LoadImage,
+  FluxKontextMultiReferenceLatentMethod set to index_timestep_zero, and
+  SamplerCustomAdvanced with KSamplerSelect euler + BasicScheduler simple/8 +
+  CFGGuider cfg 1.
+
+  8. For instruction editing (Krea 2 Identity Edit), stock nodes are not
+  enough: install the ComfyUI-Krea2Edit node pack, which ships its own
+  workflows, and run Turbo at 8-12 steps / cfg 1 (or Raw at 20 steps / cfg 3
+  for removals) with LoRA strength 1.0.
+
+  9. Low-VRAM alternative: install city96's ComfyUI-GGUF and use a community
+  GGUF build (realrebelai/KREA-2_GGUFs, quantized from Turbo;
+  molbal/krea2-gguf, quantized from Raw) with the Unet Loader (GGUF) node in
+  place of UNETLoader. There is no official GGUF release. LoRA loading is
+  unchanged.
+
+───────────────────────────────────────────────────────────────────────────────
+  Scraped by lorakit on 2026-09-05 04:00 UTC.
+  Stats and description are the author's; pros/cons are derived from the
+  evidence tagged beside each line.
+───────────────────────────────────────────────────────────────────────────────

cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b/metadata.json

@@ -0,0 +1,137 @@
+{
+  "name": "Cheerleader outfit - (Krea 2, Anima, Illustrious, Flux 2 Klein 9B, Flux 2 Klein 4B & ZIT)",
+  "slug": "cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b",
+  "base_model": "Krea 2",
+  "media_type": "image",
+  "model_id": 2610117,
+  "version_id": 3200034,
+  "version_name": "Krea 2 Version",
+  "source": "civitai",
+  "url": "https://civitai.com/models/2610117",
+  "repo_id": null,
+  "file_path": null,
+  "author": "DTAL835217",
+  "category": {
+    "key": "clothing",
+    "label": "Clothing / wardrobe"
+  },
+  "modalities": [],
+  "primary_modality": "unspecified",
+  "checkpoint_family": "unknown",
+  "trigger_words": [],
+  "recommended_settings": {
+    "LoRA strength": "0.7–1.0"
+  },
+  "chaining": {
+    "author": {
+      "stacks": "not stated",
+      "notes": "",
+      "avoid": [],
+      "requires": [],
+      "order": "",
+      "strength": "",
+      "stated": false
+    },
+    "general_rules": [
+      "The most important rule is not about chaining but about which checkpoint you chain onto: the official README says in bold \"TRAIN on Raw and RUN on Turbo\", and \"We highly recommend using RAW for training LoRAs and applying them on Turbo for inference.\" LoRAs trained on Raw are stated to work well on Turbo.",
+      "Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2 LoRAs are DiT-only, so there is no CLIP side to chain and the text-encoder branch stays untouched however many LoRAs you add.",
+      "Krea 2 LoRAs stack in practice and the community treats multi-LoRA as routine - style + character, style + slider, realism + character. Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one writes \"I recently stacked over 20 of my own LoRAs on a single generation and the output still came out great.\"",
+      "Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is loaded. The CivitAI 12GB training guide also notes rank affects composability: \"rank 8 produced a lighter style influence that was easier to stack with other LoRAs, rank 16 provided a useful balance between concept strength and composability, and rank 32 was better suited to a precise object or concept intended to take priority over other stacked LoRAs.\"",
+      "Accelerator LoRAs are position-sensitive in intent, not in node order: a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is explicitly recommended by the 4-step LoRA's author.",
+      "Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card says \"Keep mu = 1.15. The training targets are anchored to that grid; a different shift moves the [sampler off them]\"; the same reasoning applies to any Turbo-trained adapter in the chain.",
+      "Reference/edit LoRAs are designed to be stacked under ordinary subject and style LoRAs - the Identity Edit card states \"Composes with your LoRAs: character/body/style LoRAs stack on top and steer the prior\" and recommends a stacked subject LoRA when unusual subjects drift toward the base prior.",
+      "Krea 2 LoRAs cannot be chained with LoRAs for any other base model. Many CivitAI pages carry multi-model titles like \"(Krea-2 + ZIT)\" or \"[Flux | ZIB | Krea2]\" - those are separate files for separate architectures shipped under one model page, not a combined adapter.",
+      "To combine two full Krea 2 checkpoints rather than adapters, ComfyUI ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-section weights for first., tmlp., txtmlp., tproj., txtfusion.projector, txtfusion.layerwise_blocks.0-1, txtfusion.refiner_blocks.0-1, blocks.0-27 and last.",
+      "The author never states which weight family this targets. Silent degradation, never a hard error, between Raw and Turbo. The two share one architecture - ai-toolkit's config is commented \"The reference 'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)\" - so every LoRA tensor shape matches on both and ComfyUI/diffuser"
+    ]
+  },
+  "pros": [
+    "Well established — 2,602 downloads. [source: CivitAI stats]",
+    "Positively received — 127 thumbs-up with no down-votes. [source: CivitAI stats]",
+    "Compact at 218 MB — cheap to stack with other LoRAs. [source: file size]",
+    "Author published concrete recommended settings. [source: description]"
+  ],
+  "cons": [
+    "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]",
+    "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]",
+    "Thin documentation — expect trial and error. [source: description length]"
+  ],
+  "file": {
+    "name": "Cheerleader_outfit_-_Krea_2_epoch_10.safetensors",
+    "size_bytes": 228588912,
+    "size_mb": 218.0,
+    "sha256": "6f9a6764cb959fc46e10b2f142cffcbe92dfa8b5fe20b5c0abad71d98fc30980"
+  },
+  "stats": {
+    "downloads": 2602,
+    "thumbs_up": 127,
+    "thumbs_down": 0
+  },
+  "popularity": {
+    "downloads": 2602,
+    "thumbs_up": 127,
+    "thumbs_down": 0,
+    "comments": 0,
+    "approval": "100% positive (127/127)",
+    "age_days": 30.7,
+    "downloads_per_day": 27.8,
+    "rank_by_downloads": 94,
+    "rank_by_momentum": 153,
+    "of_total": 200,
+    "adoption_tier": "Moderate",
+    "momentum": "Long tail · fresh release"
+  },
+  "selection_score": 84.4,
+  "selection_reasons": [
+    "2,602 downloads",
+    "127 thumbs-up",
+    "top-ranked for its category"
+  ],
+  "example_media": [
+    {
+      "file": "01_image.jpeg",
+      "type": "image",
+      "resolution": "1024x768",
+      "prompt": ""
+    },
+    {
+      "file": "02_image.jpeg",
+      "type": "image",
+      "resolution": "832x1216",
+      "prompt": ""
+    },
+    {
+      "file": "03_image.jpeg",
+      "type": "image",
+      "resolution": "1024x768",
+      "prompt": ""
+    },
+    {
+      "file": "04_image.jpeg",
+      "type": "image",
+      "resolution": "832x1216",
+      "prompt": ""
+    },
+    {
+      "file": "05_image.jpeg",
+      "type": "image",
+      "resolution": "1024x768",
+      "prompt": ""
+    },
+    {
+      "file": "06_video.mp4",
+      "type": "video",
+      "resolution": "720x1280",
+      "prompt": "",
+      "poster": "06_poster.jpg"
+    }
+  ],
+  "tags": [
+    "outfit",
+    "cheerleader",
+    "cheerleader outfit",
+    "clothing"
+  ],
+  "scraped_at": "2026-09-05T04:00:53+00:00",
+  "path": "krea-2/cheerleader-outfit-krea-2-anima-illustrious-flux-2-klein-9b"
+}

crt-vaporwave/crt_vaporwave_krea2.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d08af66644a9df1deba5344b9a60450c4cc743b39f3af850c407736464608c7e
+size 228587792

crt-vaporwave/info.txt

@@ -0,0 +1,295 @@
+═══════════════════════════════════════════════════════════════════════════════
+  CRT VAPORWAVE
+═══════════════════════════════════════════════════════════════════════════════
+
+───────────────────────────────────────────────────────────────────────────────
+  IDENTITY
+───────────────────────────────────────────────────────────────────────────────
+  Base model        Krea 2
+  Generation mode   not declared by the author
+  Checkpoint        not determined — verify before use
+  Category          Visual style
+  Author            HackAfterDark
+  Source            https://civitai.com/models/2856432
+  Provenance        CivitAI · model 2856432 · version 3226199
+
+───────────────────────────────────────────────────────────────────────────────
+  WEIGHTS
+───────────────────────────────────────────────────────────────────────────────
+  File              crt_vaporwave_krea2.safetensors
+  Version           v1.0
+  Size              218.0 MB
+  Format            SafeTensor / unknown precision
+  SHA256            d08af66644a9df1deba5344b9a60450c4cc743b39f3af850c407736464608c7e
+  Published         2026-08-13
+
+───────────────────────────────────────────────────────────────────────────────
+  TRIGGER WORDS
+───────────────────────────────────────────────────────────────────────────────
+  › "crtvaporwave"
+
+───────────────────────────────────────────────────────────────────────────────
+  RECOMMENDED SETTINGS
+───────────────────────────────────────────────────────────────────────────────
+  Taken verbatim from the author's description:
+
+  LoRA strength       0.1–1.0
+  Sampler             euler
+
+───────────────────────────────────────────────────────────────────────────────
+  ADOPTION & POPULARITY
+───────────────────────────────────────────────────────────────────────────────
+  Downloads               598
+  Rating                  100% positive (120/120)
+  Comments                5
+  Published               22 days ago
+  Download rate           26.9/day since release
+
+  Adoption rank           #184 of 200 in this collection  (Niche)
+  Momentum rank           #158 of 200 by download rate  (Long tail · fresh release)
+
+  Adoption tier is the percentile of total downloads within this collection;
+  momentum is the percentile of downloads-per-day since release. Momentum is
+  ranked rather than measured against a fixed rate, because this base model
+  is itself new and every LoRA looks fast on an absolute scale. Both numbers
+  are CivitAI's, read on the scrape date at the foot of this file.
+
+───────────────────────────────────────────────────────────────────────────────
+  CHAINING / STACKING
+───────────────────────────────────────────────────────────────────────────────
+  Stacks with others      not stated
+
+  The author gave no chaining guidance. The rules below are the
+  general Krea 2 ones, not anything specific to this LoRA.
+
+  General Krea 2 rules that apply here (see REFERENCE.txt):
+
+  • The most important rule is not about chaining but about which
+    checkpoint you chain onto: the official README says in bold "TRAIN on
+    Raw and RUN on Turbo", and "We highly recommend using RAW for training
+    LoRAs and applying them on Turbo for inference." LoRAs trained on Raw
+    are stated to work well on Turbo.
+  • Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2
+    LoRAs are DiT-only, so there is no CLIP side to chain and the text-
+    encoder branch stays untouched however many LoRAs you add.
+  • Krea 2 LoRAs stack in practice and the community treats multi-LoRA as
+    routine - style + character, style + slider, realism + character.
+    Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one
+    writes "I recently stacked over 20 of my own LoRAs on a single
+    generation and the output still came out great."
+  • Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge
+    on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is
+    loaded. The CivitAI 12GB training guide also notes rank affects
+    composability: "rank 8 produced a lighter style influence that was
+    easier to stack with other LoRAs, rank 16 provided a useful balance
+    between concept strength and composability, and rank 32 was better
+    suited to a precise object or concept intended to take priority over
+    other stacked LoRAs."
+  • Accelerator LoRAs are position-sensitive in intent, not in node order:
+    a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a
+    sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-
+    Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is
+    explicitly recommended by the 4-step LoRA's author.
+  • Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI
+    (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card
+    says "Keep mu = 1.15. The training targets are anchored to that grid; a
+    different shift moves the [sampler off them]"; the same reasoning
+    applies to any Turbo-trained adapter in the chain.
+  • Reference/edit LoRAs are designed to be stacked under ordinary subject
+    and style LoRAs - the Identity Edit card states "Composes with your
+    LoRAs: character/body/style LoRAs stack on top and steer the prior" and
+    recommends a stacked subject LoRA when unusual subjects drift toward
+    the base prior.
+  • Krea 2 LoRAs cannot be chained with LoRAs for any other base model.
+    Many CivitAI pages carry multi-model titles like "(Krea-2 + ZIT)" or
+    "[Flux | ZIB | Krea2]" - those are separate files for separate
+    architectures shipped under one model page, not a combined adapter.
+  • To combine two full Krea 2 checkpoints rather than adapters, ComfyUI
+    ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-
+    section weights for first., tmlp., txtmlp., tproj.,
+    txtfusion.projector, txtfusion.layerwise_blocks.0-1,
+    txtfusion.refiner_blocks.0-1, blocks.0-27 and last.
+  • The author never states which weight family this targets. Silent
+    degradation, never a hard error, between Raw and Turbo. The two share
+    one architecture - ai-toolkit's config is commented "The reference
+    'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)"
+    - so every LoRA tensor shape matches on both and ComfyUI/diffuser
+
+───────────────────────────────────────────────────────────────────────────────
+  PROS
+───────────────────────────────────────────────────────────────────────────────
+  + Positively received — 120 thumbs-up with no down-votes. [source:
+    CivitAI stats]
+  + Explicit trigger word(s) — crtvaporwave — so the effect can be turned
+    on and off from the prompt. [source: model metadata]
+  + Compact at 218 MB — cheap to stack with other LoRAs. [source: file
+    size]
+  + Author published concrete recommended settings. [source: description]
+
+───────────────────────────────────────────────────────────────────────────────
+  CONS & CAVEATS
+───────────────────────────────────────────────────────────────────────────────
+  - Author never states which generation mode it was trained for; test
+    against your own workflow before relying on it. [source: description]
+  - Modest adoption (598 downloads) — less community feedback to rely on.
+    [source: source stats]
+
+───────────────────────────────────────────────────────────────────────────────
+  COMPATIBILITY WARNING
+───────────────────────────────────────────────────────────────────────────────
+  Silent degradation, never a hard error, between Raw and Turbo. The two
+  share one architecture - ai-toolkit's config is commented "The reference
+  'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)" - so
+  every LoRA tensor shape matches on both and ComfyUI/diffusers will load an
+  adapter against either without complaint. The lvladikov 4-step LoRA card
+  states this for the reverse direction: "Every layer it targets also exists
+  in Krea 2 Raw, so it will load there without complaint - but that is a side
+  effect of the shared architecture, not a supported mode", and reports
+  mixed, subject-dependent results with malformed subjects (duplicated heads,
+  fused limbs) at low step counts, needing 14+ steps with Raw's CFG on before
+  output is coherent. The intended direction (train on Raw, run on Turbo) is
+  officially blessed. The more common practical failure is a settings
+  mismatch rather than a weight mismatch: running a Turbo-targeted LoRA at
+  Raw's 52 steps + CFG 3.5, or leaving CFG enabled on Turbo, degrades output
+  with no warning. Loading a Krea 2 LoRA against a different model family
+  entirely (Flux, Qwen-Image, Z-Image) is also not an error - ComfyUI logs
+  "lora key not loaded: {key}" per key and generates as if no LoRA were
+  applied.
+
+───────────────────────────────────────────────────────────────────────────────
+  AUTHOR'S DESCRIPTION
+───────────────────────────────────────────────────────────────────────────────
+  An Analog CRT & Retro Cyberpunk Aesthetic LoRA This is a specialized style
+  LoRA trained to inject authentic analog decay, retro CRT monitor physics,
+  RGB chromatic aberration, and vintage broadcast interference into your
+  generations.
+  Unlike clean digital render models, this checkpoint transforms crisp
+  digital assets into gritty, atmospheric photographs of physical CRT
+  displays and analog media.
+  Suggested Krea 2 Turbo Settings Model strength: 0.1-1.0
+  Samplers/schedulers:
+  - dpm_2_ancestral / ddim_uniform (or sgm_uniform )
+
+  - exp_heun_2_x0 / ddim_uniform
+
+  - euler / beta57 (or beta )
+
+  Also note that you can push the model strength to 1.25 as well. I think
+  beyond that gets a little too distorted, but try a variety of LoRA strength
+  values! You can really dial things in to taste.
+  📺 What It Does Best - Integrated Hardware Decay: Adds authentic horizontal
+  scanline tearing, phosphor glow bleed, and rolling electromagnetic
+  interference.
+
+  - Color Fringing: Generates organic neon cyan and magenta RGB ghosting and
+  chromatic aberration.
+
+  - Text & UI Fidelity: When dialed to the sweet spot (0.6–0.8 strength), it
+  preserves complex typography and UI elements while wrapping them in raw
+  retro atmosphere.
+
+  While Krea 2 base model does understand the elements of this style, the
+  effects it applies look a bit "pasted on top" in a fake way. This LoRA
+  integrates those effects into the image. Krea 2 by itself also doesn't do a
+  good job with the glow or haze from an old CRT screen, it's often missing.
+  This LoRA corrects the overly digital, sterile, and bright look from the
+  base model.
+  💡 Advanced Tip: Negative Strength ( -1.0 ) Want a completely different
+  look? Try setting the LoRA weight to a negative value like -1.0 .
+  Instead of adding analog CRT decay, inverting the weight acts as an "anti-
+  CRT" filter that strips away the style bias. It forces the base model to
+  pivot toward hyper-detailed, clean chrome/glass material textures and
+  sharper structural geometry while keeping the chaotic glitch borders
+  intact.
+
+───────────────────────────────────────────────────────────────────────────────
+  VERSION NOTES
+───────────────────────────────────────────────────────────────────────────────
+  This version used a batch size of 1 and was trained to 3000 steps. It was a
+  tough choice between this and another version that used an effective batch
+  size of 16 trained to 2000 steps. This version was working a little better,
+  especially when applied at higher strengths.
+
+───────────────────────────────────────────────────────────────────────────────
+  EXAMPLE MEDIA
+───────────────────────────────────────────────────────────────────────────────
+  6 preview(s) in ./example_images/ — community results for
+  this LoRA. Videos keep a matching _poster.jpg still frame.
+
+  No prompts are recorded below: CivitAI's public API flags these posts as
+  having generation metadata but does not return it, so the prompts cannot be
+  scraped. Open the model page to read them.
+
+  01_image.jpeg  (image, 1080x1920)
+  02_image.jpeg  (image, 2512x3344)
+  03_image.jpeg  (image, 832x1248)
+  04_image.jpeg  (image, 832x1248)
+  05_image.jpeg  (image, 832x1248)
+  06_image.jpeg  (image, 832x1248)
+
+───────────────────────────────────────────────────────────────────────────────
+  HOW TO USE (COMFYUI)
+───────────────────────────────────────────────────────────────────────────────
+  1. Update ComfyUI to at least v0.26.0 (released 2026-06-23), which added
+  local Krea2 support. For reference-image / style-reference / edit LoRAs you
+  need at least v0.29.0 (2026-07-29), which added regular and timestep-zero
+  reference images for Krea 2. ModelMergeKrea2 arrived in v0.27.0
+  (2026-06-30).
+
+  2. Put the base stack in place (files from huggingface.co/Comfy-Org/Krea-2,
+  which is not gated):
+     - ComfyUI/models/diffusion_models/krea2_turbo_fp8_scaled.safetensors
+  (documented recommended default; the repo also holds krea2_turbo_bf16,
+  _int8_convrot, _mxfp8, _nvfp4 and krea2_raw_bf16 / _fp8_scaled /
+  _int8_convrot)
+     - ComfyUI/models/text_encoders/qwen3vl_4b_fp8_scaled.safetensors (bf16
+  also available)
+     - ComfyUI/models/vae/qwen_image_vae.safetensors
+
+  3. Put the LoRA .safetensors in ComfyUI/models/loras/ - flat, no per-model
+  subfolder required.
+
+  4. Build or load the graph. The official image_krea2_turbo_t2i.json
+  template is: UNETLoader (krea2_turbo_fp8_scaled.safetensors) ->
+  LoraLoaderModelOnly -> KSampler; CLIPLoader
+  (qwen3vl_4b_fp8_scaled.safetensors, type "krea2") -> text encode ->
+  KSampler positive, with ConditioningZeroOut into negative; VAELoader
+  (qwen_image_vae.safetensors) -> VAEDecode -> SaveImage; EmptyLatentImage
+  1024x1024 driven by ResolutionSelector. An int8 sibling template
+  (image_krea2_turbo_t2i_int8.json) also ships.
+
+  5. Load the LoRA with LoraLoaderModelOnly, not LoraLoader - both official
+  templates do. Krea 2 LoRAs patch the DiT only; note that the DiT's own
+  text_fusion / txtfusion blocks are a normal LoRA target (the official style
+  LoRAs patch them), but the Qwen3-VL text encoder itself is not. Set
+  strength 1.0 for Krea's official style LoRAs (the shipped template uses
+  0.8) and add the LoRA's trigger word to the prompt.
+
+  6. Sampler settings: steps 8, cfg 1, sampler euler, scheduler simple,
+  denoise 1.0. Do not add a ModelSampling node for plain t2i - ComfyUI
+  already applies shift 1.15 from Krea2.sampling_settings.
+
+  7. For style reference, use the dedicated template: UNETLoader
+  krea2_turbo_int8_convrot.safetensors, LoraLoaderModelOnly
+  krea2_style_reference.safetensors at 1.0, ModelSamplingFlux (1.15, 0.5),
+  TextEncodeQwenImageEditPlus fed by LoadImage,
+  FluxKontextMultiReferenceLatentMethod set to index_timestep_zero, and
+  SamplerCustomAdvanced with KSamplerSelect euler + BasicScheduler simple/8 +
+  CFGGuider cfg 1.
+
+  8. For instruction editing (Krea 2 Identity Edit), stock nodes are not
+  enough: install the ComfyUI-Krea2Edit node pack, which ships its own
+  workflows, and run Turbo at 8-12 steps / cfg 1 (or Raw at 20 steps / cfg 3
+  for removals) with LoRA strength 1.0.
+
+  9. Low-VRAM alternative: install city96's ComfyUI-GGUF and use a community
+  GGUF build (realrebelai/KREA-2_GGUFs, quantized from Turbo;
+  molbal/krea2-gguf, quantized from Raw) with the Unet Loader (GGUF) node in
+  place of UNETLoader. There is no official GGUF release. LoRA loading is
+  unchanged.
+
+───────────────────────────────────────────────────────────────────────────────
+  Scraped by lorakit on 2026-09-05 04:00 UTC.
+  Stats and description are the author's; pros/cons are derived from the
+  evidence tagged beside each line.
+───────────────────────────────────────────────────────────────────────────────

crt-vaporwave/metadata.json

@@ -0,0 +1,142 @@
+{
+  "name": "CRT Vaporwave",
+  "slug": "crt-vaporwave",
+  "base_model": "Krea 2",
+  "media_type": "image",
+  "model_id": 2856432,
+  "version_id": 3226199,
+  "version_name": "v1.0",
+  "source": "civitai",
+  "url": "https://civitai.com/models/2856432",
+  "repo_id": null,
+  "file_path": null,
+  "author": "HackAfterDark",
+  "category": {
+    "key": "style",
+    "label": "Visual style"
+  },
+  "modalities": [],
+  "primary_modality": "unspecified",
+  "checkpoint_family": "unknown",
+  "trigger_words": [
+    "crtvaporwave"
+  ],
+  "recommended_settings": {
+    "LoRA strength": "0.1–1.0",
+    "Sampler": "euler"
+  },
+  "chaining": {
+    "author": {
+      "stacks": "not stated",
+      "notes": "",
+      "avoid": [],
+      "requires": [],
+      "order": "",
+      "strength": "",
+      "stated": false
+    },
+    "general_rules": [
+      "The most important rule is not about chaining but about which checkpoint you chain onto: the official README says in bold \"TRAIN on Raw and RUN on Turbo\", and \"We highly recommend using RAW for training LoRAs and applying them on Turbo for inference.\" LoRAs trained on Raw are stated to work well on Turbo.",
+      "Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2 LoRAs are DiT-only, so there is no CLIP side to chain and the text-encoder branch stays untouched however many LoRAs you add.",
+      "Krea 2 LoRAs stack in practice and the community treats multi-LoRA as routine - style + character, style + slider, realism + character. Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one writes \"I recently stacked over 20 of my own LoRAs on a single generation and the output still came out great.\"",
+      "Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is loaded. The CivitAI 12GB training guide also notes rank affects composability: \"rank 8 produced a lighter style influence that was easier to stack with other LoRAs, rank 16 provided a useful balance between concept strength and composability, and rank 32 was better suited to a precise object or concept intended to take priority over other stacked LoRAs.\"",
+      "Accelerator LoRAs are position-sensitive in intent, not in node order: a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is explicitly recommended by the 4-step LoRA's author.",
+      "Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card says \"Keep mu = 1.15. The training targets are anchored to that grid; a different shift moves the [sampler off them]\"; the same reasoning applies to any Turbo-trained adapter in the chain.",
+      "Reference/edit LoRAs are designed to be stacked under ordinary subject and style LoRAs - the Identity Edit card states \"Composes with your LoRAs: character/body/style LoRAs stack on top and steer the prior\" and recommends a stacked subject LoRA when unusual subjects drift toward the base prior.",
+      "Krea 2 LoRAs cannot be chained with LoRAs for any other base model. Many CivitAI pages carry multi-model titles like \"(Krea-2 + ZIT)\" or \"[Flux | ZIB | Krea2]\" - those are separate files for separate architectures shipped under one model page, not a combined adapter.",
+      "To combine two full Krea 2 checkpoints rather than adapters, ComfyUI ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-section weights for first., tmlp., txtmlp., tproj., txtfusion.projector, txtfusion.layerwise_blocks.0-1, txtfusion.refiner_blocks.0-1, blocks.0-27 and last.",
+      "The author never states which weight family this targets. Silent degradation, never a hard error, between Raw and Turbo. The two share one architecture - ai-toolkit's config is commented \"The reference 'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)\" - so every LoRA tensor shape matches on both and ComfyUI/diffuser"
+    ]
+  },
+  "pros": [
+    "Positively received — 120 thumbs-up with no down-votes. [source: CivitAI stats]",
+    "Explicit trigger word(s) — crtvaporwave — so the effect can be turned on and off from the prompt. [source: model metadata]",
+    "Compact at 218 MB — cheap to stack with other LoRAs. [source: file size]",
+    "Author published concrete recommended settings. [source: description]"
+  ],
+  "cons": [
+    "Author never states which generation mode it was trained for; test against your own workflow before relying on it. [source: description]",
+    "Modest adoption (598 downloads) — less community feedback to rely on. [source: source stats]"
+  ],
+  "file": {
+    "name": "crt_vaporwave_krea2.safetensors",
+    "size_bytes": 228587792,
+    "size_mb": 218.0,
+    "sha256": "d08af66644a9df1deba5344b9a60450c4cc743b39f3af850c407736464608c7e"
+  },
+  "stats": {
+    "downloads": 598,
+    "thumbs_up": 120,
+    "thumbs_down": 0
+  },
+  "popularity": {
+    "downloads": 598,
+    "thumbs_up": 120,
+    "thumbs_down": 0,
+    "comments": 5,
+    "approval": "100% positive (120/120)",
+    "age_days": 22.2,
+    "downloads_per_day": 26.91,
+    "rank_by_downloads": 184,
+    "rank_by_momentum": 158,
+    "of_total": 200,
+    "adoption_tier": "Niche",
+    "momentum": "Long tail · fresh release"
+  },
+  "selection_score": 84.7,
+  "selection_reasons": [
+    "598 downloads",
+    "120 thumbs-up",
+    "documented triggers + usage notes",
+    "top-ranked for its category"
+  ],
+  "example_media": [
+    {
+      "file": "01_image.jpeg",
+      "type": "image",
+      "resolution": "1080x1920",
+      "prompt": ""
+    },
+    {
+      "file": "02_image.jpeg",
+      "type": "image",
+      "resolution": "2512x3344",
+      "prompt": ""
+    },
+    {
+      "file": "03_image.jpeg",
+      "type": "image",
+      "resolution": "832x1248",
+      "prompt": ""
+    },
+    {
+      "file": "04_image.jpeg",
+      "type": "image",
+      "resolution": "832x1248",
+      "prompt": ""
+    },
+    {
+      "file": "05_image.jpeg",
+      "type": "image",
+      "resolution": "832x1248",
+      "prompt": ""
+    },
+    {
+      "file": "06_image.jpeg",
+      "type": "image",
+      "resolution": "832x1248",
+      "prompt": ""
+    }
+  ],
+  "tags": [
+    "crt",
+    "style",
+    "glitch",
+    "vhs",
+    "cyber",
+    "retro",
+    "vaporwave"
+  ],
+  "scraped_at": "2026-09-05T04:00:29+00:00",
+  "path": "krea-2/crt-vaporwave"
+}

dgirls-multiple-females/DGirlsModel3.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:186c4348239193ec3941f0e154759d8b0de848a3bf627caaac0a80a12a8f4569
+size 228587784

dgirls-multiple-females/info.txt

@@ -0,0 +1,305 @@
+═══════════════════════════════════════════════════════════════════════════════
+  DGIRLS - MULTIPLE FEMALES
+═══════════════════════════════════════════════════════════════════════════════
+
+───────────────────────────────────────────────────────────────────────────────
+  IDENTITY
+───────────────────────────────────────────────────────────────────────────────
+  Base model        Krea 2
+  Generation mode   Instruction / reference editing (community, not official)
+  Checkpoint        Krea 2 RAW (krea/Krea-2-Raw)
+  Category          Photorealism
+  Author            DGirls
+  Source            https://civitai.com/models/2911848
+  Provenance        CivitAI · model 2911848 · version 3296360
+
+───────────────────────────────────────────────────────────────────────────────
+  WEIGHTS
+───────────────────────────────────────────────────────────────────────────────
+  File              DGirlsModel3.safetensors
+  Version           V3.0 Ruby
+  Size              218.0 MB
+  Format            SafeTensor / unknown precision
+  SHA256            186c4348239193ec3941f0e154759d8b0de848a3bf627caaac0a80a12a8f4569
+  Published         2026-09-04
+
+───────────────────────────────────────────────────────────────────────────────
+  TRIGGER WORDS
+───────────────────────────────────────────────────────────────────────────────
+  No trigger word. The LoRA applies as soon as it is loaded —
+  strength is the only way to dial the effect up or down.
+
+───────────────────────────────────────────────────────────────────────────────
+  RECOMMENDED SETTINGS
+───────────────────────────────────────────────────────────────────────────────
+  The author did not publish specific settings.
+  No author settings were published. Community baseline: steps Turbo: 8
+  (official README and HF card CLI example; the ComfyUI template ships 8).
+  Raw: 52 (official README). The krea-2 CLI's generic default is 28.
+  Community Turbo LoRAs commonly suggest 8-12 (e.g. the Identity Edit card
+  recommends Turbo at 8-12, or Raw at 20 for removals). A community 4-step
+  LoRA on Turbo makes 4 valid; its author calls 2 steps an out-of-spec
+  preview mode., cfg Turbo: CFG disabled - guidance_scale 0.0 in diffusers
+  and --cfg 0.0 in the krea-2 CLI, which is cfg 1.0 in ComfyUI's KSampler
+  (the official template ships cfg 1). These are the same thing; do not set
+  ComfyUI cfg to 0. Raw: CFG 3.5 (official README; CLI generic default 4.5).
+  Negative prompting on Turbo is effectively unused - the official templates
+  route ConditioningZeroOut into the negative slot., sampler ComfyUI:
+  KSampler with sampler euler, scheduler simple, denoise 1.0 (verified in
+  image_krea2_turbo_t2i.json: [seed, 'randomize', 8, 1, 'euler', 'simple',
+  1]). Timestep shift mu = 1.15 for Turbo; ComfyUI bakes this in via
+  Krea2.sampling_settings = {"multiplier": 1.0, "shift": 1.15}, so the plain
+  t2i template needs no ModelSampling node, while the style-reference
+  template adds ModelSamplingFlux at max_shift 1.15 / base_shift 0.5. The CLI
+  exposes --mu 1.15 for Turbo; Raw uses a resolution-derived mu interpolated
+  between --y1 0.5 and --y2 1.15 (ai-toolkit mirrors this as base_shift 0.5 /
+  max_shift 1.15 over image_seq_len 256-6400, exponential time shift).,
+  resolution 1024x1024 default. Turbo: 1K-2K (the official CLI example
+  generates 2048x2048; ComfyUI's ResolutionSelector defaults to 1 megapixel,
+  set 2.0 for 2K). Raw: "trained to generate upto 1k resolution".
+  Width/height are padded up to a multiple of 16., lora_strength 1.0 for
+  Krea's nine official style LoRAs (docs.comfy.org lists Recommended Strength
+  1.0 for every one), though the shipped t2i template's LoraLoaderModelOnly
+  is set to 0.8. The style-reference LoRA is used at 1.0 in its template. The
+  4-step distill LoRA documents 1.0 as its trained point, 1.0-1.5 as coherent
+  extrapolation and above ~1.5 as speckle artefacting. Community subject
+  LoRAs are typically recommended around 0.8-1.2 and dropped toward 0.6-0.9
+  when several are stacked..
+
+───────────────────────────────────────────────────────────────────────────────
+  ADOPTION & POPULARITY
+───────────────────────────────────────────────────────────────────────────────
+  Downloads               479
+  Rating                  100% positive (19/19)
+  Comments                0
+  Published               0 days ago
+  Download rate           83.0/day since release
+
+  Adoption rank           #196 of 200 in this collection  (Niche)
+  Momentum rank           #37 of 200 by download rate  (Trending · fresh release)
+
+  Adoption tier is the percentile of total downloads within this collection;
+  momentum is the percentile of downloads-per-day since release. Momentum is
+  ranked rather than measured against a fixed rate, because this base model
+  is itself new and every LoRA looks fast on an absolute scale. Both numbers
+  are CivitAI's, read on the scrape date at the foot of this file.
+
+───────────────────────────────────────────────────────────────────────────────
+  CHAINING / STACKING
+───────────────────────────────────────────────────────────────────────────────
+  Stacks with others      not stated
+
+  The author gave no chaining guidance. The rules below are the
+  general Krea 2 ones, not anything specific to this LoRA.
+
+  General Krea 2 rules that apply here (see REFERENCE.txt):
+
+  • The most important rule is not about chaining but about which
+    checkpoint you chain onto: the official README says in bold "TRAIN on
+    Raw and RUN on Turbo", and "We highly recommend using RAW for training
+    LoRAs and applying them on Turbo for inference." LoRAs trained on Raw
+    are stated to work well on Turbo.
+  • Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2
+    LoRAs are DiT-only, so there is no CLIP side to chain and the text-
+    encoder branch stays untouched however many LoRAs you add.
+  • Krea 2 LoRAs stack in practice and the community treats multi-LoRA as
+    routine - style + character, style + slider, realism + character.
+    Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one
+    writes "I recently stacked over 20 of my own LoRAs on a single
+    generation and the output still came out great."
+  • Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge
+    on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is
+    loaded. The CivitAI 12GB training guide also notes rank affects
+    composability: "rank 8 produced a lighter style influence that was
+    easier to stack with other LoRAs, rank 16 provided a useful balance
+    between concept strength and composability, and rank 32 was better
+    suited to a precise object or concept intended to take priority over
+    other stacked LoRAs."
+  • Accelerator LoRAs are position-sensitive in intent, not in node order:
+    a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a
+    sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-
+    Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is
+    explicitly recommended by the 4-step LoRA's author.
+  • Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI
+    (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card
+    says "Keep mu = 1.15. The training targets are anchored to that grid; a
+    different shift moves the [sampler off them]"; the same reasoning
+    applies to any Turbo-trained adapter in the chain.
+  • Reference/edit LoRAs are designed to be stacked under ordinary subject
+    and style LoRAs - the Identity Edit card states "Composes with your
+    LoRAs: character/body/style LoRAs stack on top and steer the prior" and
+    recommends a stacked subject LoRA when unusual subjects drift toward
+    the base prior.
+  • Krea 2 LoRAs cannot be chained with LoRAs for any other base model.
+    Many CivitAI pages carry multi-model titles like "(Krea-2 + ZIT)" or
+    "[Flux | ZIB | Krea2]" - those are separate files for separate
+    architectures shipped under one model page, not a combined adapter.
+  • To combine two full Krea 2 checkpoints rather than adapters, ComfyUI
+    ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-
+    section weights for first., tmlp., txtmlp., tproj.,
+    txtfusion.projector, txtfusion.layerwise_blocks.0-1,
+    txtfusion.refiner_blocks.0-1, blocks.0-27 and last.
+  • Undistilled base checkpoint, no step or CFG distillation and no post-
+    training. The card says "its not recommended for inference use. Its a
+    good base for finetuning or post-training". Official settings: 52
+    steps, CFG 3.5, trained up to 1K. This is where Krea tells you to train
+    LoRAs.
+
+───────────────────────────────────────────────────────────────────────────────
+  PROS
+───────────────────────────────────────────────────────────────────────────────
+  + Compact at 218 MB — cheap to stack with other LoRAs. [source: file
+    size]
+
+───────────────────────────────────────────────────────────────────────────────
+  CONS & CAVEATS
+───────────────────────────────────────────────────────────────────────────────
+  - No trigger word — the LoRA is always on once loaded, so strength is the
+    only control. [source: model metadata]
+  - Modest adoption (479 downloads) — less community feedback to rely on.
+    [source: source stats]
+
+───────────────────────────────────────────────────────────────────────────────
+  COMPATIBILITY WARNING
+───────────────────────────────────────────────────────────────────────────────
+  Silent degradation, never a hard error, between Raw and Turbo. The two
+  share one architecture - ai-toolkit's config is commented "The reference
+  'single_mmdit_large_wide' architecture (oss_raw / oss_turbo share it)" - so
+  every LoRA tensor shape matches on both and ComfyUI/diffusers will load an
+  adapter against either without complaint. The lvladikov 4-step LoRA card
+  states this for the reverse direction: "Every layer it targets also exists
+  in Krea 2 Raw, so it will load there without complaint - but that is a side
+  effect of the shared architecture, not a supported mode", and reports
+  mixed, subject-dependent results with malformed subjects (duplicated heads,
+  fused limbs) at low step counts, needing 14+ steps with Raw's CFG on before
+  output is coherent. The intended direction (train on Raw, run on Turbo) is
+  officially blessed. The more common practical failure is a settings
+  mismatch rather than a weight mismatch: running a Turbo-targeted LoRA at
+  Raw's 52 steps + CFG 3.5, or leaving CFG enabled on Turbo, degrades output
+  with no warning. Loading a Krea 2 LoRA against a different model family
+  entirely (Flux, Qwen-Image, Z-Image) is also not an error - ComfyUI logs
+  "lora key not loaded: {key}" per key and generates as if no LoRA were
+  applied.
+
+───────────────────────────────────────────────────────────────────────────────
+  AUTHOR'S DESCRIPTION
+───────────────────────────────────────────────────────────────────────────────
+  🚨 NEW — RDigitalArtist LoRA HUB 🚨 📲 TELEGRAM: t.me/RDigitalArtistDGirls
+  All of my LoRA models from now on will be released through my new Telegram
+  group!
+  If you enjoy my models and want to make sure you never miss a new release ,
+  join the group and follow along. New models, updates, previews and future
+  projects will be posted there.
+  👉 JOIN THE TELEGRAM GROUP: t.me/RDigitalArtistDGirls
+  🔥 WANT YOUR OWN PRIVATE LoRA? I'm also offering FREE custom LoRA creation
+  using my LoRA training server.
+  Want a LoRA based on yourself, your own character, an OC, or a specific
+  visual identity? Send me your training images and I'll create a custom
+  model for you.
+  🔒 PRIVATE LoRA CREATION Your LoRA can be created specifically for your own
+  private use rather than being released publicly.
+  You provide the images → I handle the LoRA training → you receive your
+  custom model.
+  Perfect for creating consistent characters and identities across your AI-
+  generated images.
+  🆓 FREE TRAINING I'm currently offering the training service completely free
+  while I build and develop the service.
+  If you've been wanting to experiment with your own LoRA but don't have the
+  hardware, this is a great opportunity to give it a try.
+  📩 CONTACT ME Discord: damien2785 Telegram: DamoG2785
+  You can contact me on either platform to discuss your LoRA and submit your
+  training images.
+  Not sure whether your images will work? Message me first and I'll let you
+  know what you'll need.
+  ⭐ DON'T MISS MY NEW MODELS My future LoRA releases will be posted in the
+  Telegram group first.
+  📲 Join here: t.me/RDigitalArtistDGirls
+  Come join the RDigitalArtist / DGirls community and follow the development
+  of my upcoming LoRA models and projects.
+  Join the group → Get the models → Create your own. 🔥
+
+───────────────────────────────────────────────────────────────────────────────
+  VERSION NOTES
+───────────────────────────────────────────────────────────────────────────────
+  Model Number 3 with red hair and a Elsa type braid tuned it with torpedo
+  shape breasts
+
+───────────────────────────────────────────────────────────────────────────────
+  EXAMPLE MEDIA
+───────────────────────────────────────────────────────────────────────────────
+  2 preview(s) in ./example_images/ — community results for
+  this LoRA. Videos keep a matching _poster.jpg still frame.
+
+  No prompts are recorded below: CivitAI's public API flags these posts as
+  having generation metadata but does not return it, so the prompts cannot be
+  scraped. Open the model page to read them.
+
+  01_image.jpeg  (image, 1776x2368)
+  02_image.jpeg  (image, 1776x2368)
+
+───────────────────────────────────────────────────────────────────────────────
+  HOW TO USE (COMFYUI)
+───────────────────────────────────────────────────────────────────────────────
+  1. Update ComfyUI to at least v0.26.0 (released 2026-06-23), which added
+  local Krea2 support. For reference-image / style-reference / edit LoRAs you
+  need at least v0.29.0 (2026-07-29), which added regular and timestep-zero
+  reference images for Krea 2. ModelMergeKrea2 arrived in v0.27.0
+  (2026-06-30).
+
+  2. Put the base stack in place (files from huggingface.co/Comfy-Org/Krea-2,
+  which is not gated):
+     - ComfyUI/models/diffusion_models/krea2_turbo_fp8_scaled.safetensors
+  (documented recommended default; the repo also holds krea2_turbo_bf16,
+  _int8_convrot, _mxfp8, _nvfp4 and krea2_raw_bf16 / _fp8_scaled /
+  _int8_convrot)
+     - ComfyUI/models/text_encoders/qwen3vl_4b_fp8_scaled.safetensors (bf16
+  also available)
+     - ComfyUI/models/vae/qwen_image_vae.safetensors
+
+  3. Put the LoRA .safetensors in ComfyUI/models/loras/ - flat, no per-model
+  subfolder required.
+
+  4. Build or load the graph. The official image_krea2_turbo_t2i.json
+  template is: UNETLoader (krea2_turbo_fp8_scaled.safetensors) ->
+  LoraLoaderModelOnly -> KSampler; CLIPLoader
+  (qwen3vl_4b_fp8_scaled.safetensors, type "krea2") -> text encode ->
+  KSampler positive, with ConditioningZeroOut into negative; VAELoader
+  (qwen_image_vae.safetensors) -> VAEDecode -> SaveImage; EmptyLatentImage
+  1024x1024 driven by ResolutionSelector. An int8 sibling template
+  (image_krea2_turbo_t2i_int8.json) also ships.
+
+  5. Load the LoRA with LoraLoaderModelOnly, not LoraLoader - both official
+  templates do. Krea 2 LoRAs patch the DiT only; note that the DiT's own
+  text_fusion / txtfusion blocks are a normal LoRA target (the official style
+  LoRAs patch them), but the Qwen3-VL text encoder itself is not. Set
+  strength 1.0 for Krea's official style LoRAs (the shipped template uses
+  0.8) and add the LoRA's trigger word to the prompt.
+
+  6. Sampler settings: steps 8, cfg 1, sampler euler, scheduler simple,
+  denoise 1.0. Do not add a ModelSampling node for plain t2i - ComfyUI
+  already applies shift 1.15 from Krea2.sampling_settings.
+
+  7. For style reference, use the dedicated template: UNETLoader
+  krea2_turbo_int8_convrot.safetensors, LoraLoaderModelOnly
+  krea2_style_reference.safetensors at 1.0, ModelSamplingFlux (1.15, 0.5),
+  TextEncodeQwenImageEditPlus fed by LoadImage,
+  FluxKontextMultiReferenceLatentMethod set to index_timestep_zero, and
+  SamplerCustomAdvanced with KSamplerSelect euler + BasicScheduler simple/8 +
+  CFGGuider cfg 1.
+
+  8. For instruction editing (Krea 2 Identity Edit), stock nodes are not
+  enough: install the ComfyUI-Krea2Edit node pack, which ships its own
+  workflows, and run Turbo at 8-12 steps / cfg 1 (or Raw at 20 steps / cfg 3
+  for removals) with LoRA strength 1.0.
+
+  9. Low-VRAM alternative: install city96's ComfyUI-GGUF and use a community
+  GGUF build (realrebelai/KREA-2_GGUFs, quantized from Turbo;
+  molbal/krea2-gguf, quantized from Raw) with the Unet Loader (GGUF) node in
+  place of UNETLoader. There is no official GGUF release. LoRA loading is
+  unchanged.
+
+───────────────────────────────────────────────────────────────────────────────
+  Scraped by lorakit on 2026-09-05 04:01 UTC.
+  Stats and description are the author's; pros/cons are derived from the
+  evidence tagged beside each line.
+───────────────────────────────────────────────────────────────────────────────

dgirls-multiple-females/metadata.json

@@ -0,0 +1,111 @@
+{
+  "name": "DGirls - Multiple Females",
+  "slug": "dgirls-multiple-females",
+  "base_model": "Krea 2",
+  "media_type": "image",
+  "model_id": 2911848,
+  "version_id": 3296360,
+  "version_name": "V3.0 Ruby",
+  "source": "civitai",
+  "url": "https://civitai.com/models/2911848",
+  "repo_id": null,
+  "file_path": null,
+  "author": "DGirls",
+  "category": {
+    "key": "photoreal",
+    "label": "Photorealism"
+  },
+  "modalities": [
+    "edit"
+  ],
+  "primary_modality": "edit",
+  "checkpoint_family": "raw",
+  "trigger_words": [],
+  "recommended_settings": {},
+  "chaining": {
+    "author": {
+      "stacks": "not stated",
+      "notes": "",
+      "avoid": [],
+      "requires": [],
+      "order": "",
+      "strength": "",
+      "stated": false
+    },
+    "general_rules": [
+      "The most important rule is not about chaining but about which checkpoint you chain onto: the official README says in bold \"TRAIN on Raw and RUN on Turbo\", and \"We highly recommend using RAW for training LoRAs and applying them on Turbo for inference.\" LoRAs trained on Raw are stated to work well on Turbo.",
+      "Chain with LoraLoaderModelOnly nodes in series (model -> model). Krea 2 LoRAs are DiT-only, so there is no CLIP side to chain and the text-encoder branch stays untouched however many LoRAs you add.",
+      "Krea 2 LoRAs stack in practice and the community treats multi-LoRA as routine - style + character, style + slider, realism + character. Several CivitAI Krea 2 authors state their LoRAs stack cleanly; one writes \"I recently stacked over 20 of my own LoRAs on a single generation and the output still came out great.\"",
+      "Reduce per-LoRA strength when stacking. CivitAI Krea 2 authors converge on dropping from ~1.0 to roughly 0.6-0.9 once more than one adapter is loaded. The CivitAI 12GB training guide also notes rank affects composability: \"rank 8 produced a lighter style influence that was easier to stack with other LoRAs, rank 16 provided a useful balance between concept strength and composability, and rank 32 was better suited to a precise object or concept intended to take priority over other stacked LoRAs.\"",
+      "Accelerator LoRAs are position-sensitive in intent, not in node order: a Raw-to-Turbo delta adapter belongs on the Raw checkpoint, a sub-8-step LoRA belongs on Turbo (or on Raw already carrying a Raw-to-Turbo adapter). Chaining Raw -> Raw-to-Turbo adapter -> 4-step LoRA is explicitly recommended by the 4-step LoRA's author.",
+      "Whatever you stack, the schedule must stay Turbo's: cfg 1.0 in ComfyUI (guidance 0.0 in diffusers) and mu/shift 1.15. The 4-step LoRA's card says \"Keep mu = 1.15. The training targets are anchored to that grid; a different shift moves the [sampler off them]\"; the same reasoning applies to any Turbo-trained adapter in the chain.",
+      "Reference/edit LoRAs are designed to be stacked under ordinary subject and style LoRAs - the Identity Edit card states \"Composes with your LoRAs: character/body/style LoRAs stack on top and steer the prior\" and recommends a stacked subject LoRA when unusual subjects drift toward the base prior.",
+      "Krea 2 LoRAs cannot be chained with LoRAs for any other base model. Many CivitAI pages carry multi-model titles like \"(Krea-2 + ZIT)\" or \"[Flux | ZIB | Krea2]\" - those are separate files for separate architectures shipped under one model page, not a combined adapter.",
+      "To combine two full Krea 2 checkpoints rather than adapters, ComfyUI ships a dedicated ModelMergeKrea2 node (added in v0.27.0) with per-section weights for first., tmlp., txtmlp., tproj., txtfusion.projector, txtfusion.layerwise_blocks.0-1, txtfusion.refiner_blocks.0-1, blocks.0-27 and last.",
+      "Undistilled base checkpoint, no step or CFG distillation and no post-training. The card says \"its not recommended for inference use. Its a good base for finetuning or post-training\". Official settings: 52 steps, CFG 3.5, trained up to 1K. This is where Krea tells you to train LoRAs."
+    ]
+  },
+  "pros": [
+    "Compact at 218 MB — cheap to stack with other LoRAs. [source: file size]"
+  ],
+  "cons": [
+    "No trigger word — the LoRA is always on once loaded, so strength is the only control. [source: model metadata]",
+    "Modest adoption (479 downloads) — less community feedback to rely on. [source: source stats]"
+  ],
+  "file": {
+    "name": "DGirlsModel3.safetensors",
+    "size_bytes": 228587784,
+    "size_mb": 218.0,
+    "sha256": "186c4348239193ec3941f0e154759d8b0de848a3bf627caaac0a80a12a8f4569"
+  },
+  "stats": {
+    "downloads": 479,
+    "thumbs_up": 19,
+    "thumbs_down": 0
+  },
+  "popularity": {
+    "downloads": 479,
+    "thumbs_up": 19,
+    "thumbs_down": 0,
+    "comments": 0,
+    "approval": "100% positive (19/19)",
+    "age_days": 0.4,
+    "downloads_per_day": 83.0,
+    "rank_by_downloads": 196,
+    "rank_by_momentum": 37,
+    "of_total": 200,
+    "adoption_tier": "Niche",
+    "momentum": "Trending · fresh release"
+  },
+  "selection_score": 84.1,
+  "selection_reasons": [
+    "479 downloads",
+    "19 thumbs-up",
+    "top-ranked for its category"
+  ],
+  "example_media": [
+    {
+      "file": "01_image.jpeg",
+      "type": "image",
+      "resolution": "1776x2368",
+      "prompt": ""
+    },
+    {
+      "file": "02_image.jpeg",
+      "type": "image",
+      "resolution": "1776x2368",
+      "prompt": ""
+    }
+  ],
+  "tags": [
+    "photorealistic",
+    "character",
+    "goth",
+    "glamour",
+    "adult woman",
+    "woman",
+    "female"
+  ],
+  "scraped_at": "2026-09-05T04:01:13+00:00",
+  "path": "krea-2/dgirls-multiple-females"
+}