Add Anime Style / Gurren Lagann Lora - LTX 2.3

.gitattributes

@@ -364,3 +364,15 @@ ltx-2-3-mila/example_images/04_video.mp4 filter=lfs diff=lfs merge=lfs -text
 ltx-2-3-mila/example_images/05_video.mp4 filter=lfs diff=lfs merge=lfs -text
 ltx-2-3-mila/example_images/06_poster.jpg filter=lfs diff=lfs merge=lfs -text
 ltx-2-3-mila/example_images/06_video.mp4 filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/01_poster.jpg filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/01_video.mp4 filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/02_poster.jpg filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/02_video.mp4 filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/03_poster.jpg filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/03_video.mp4 filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/04_poster.jpg filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/04_video.mp4 filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/05_poster.jpg filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/05_video.mp4 filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/06_poster.jpg filter=lfs diff=lfs merge=lfs -text
+anime-style-gurren-lagann-lora-ltx-2-3/example_images/06_video.mp4 filter=lfs diff=lfs merge=lfs -text

anime-style-gurren-lagann-lora-ltx-2-3/example_images/01_video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b97232ae8fa27f8a4c17b5494e5eb527397b27b12678a9d35eb6ca69e775cfe
+size 1478810

anime-style-gurren-lagann-lora-ltx-2-3/example_images/02_video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef63dc95cc9c055ba3b262ed0a55eedf4ba56b9f24648d5224303eaed7cd862b
+size 2531911

anime-style-gurren-lagann-lora-ltx-2-3/example_images/03_video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a07ba6047a528ac6335f89eaf40c4eed191c5e8409989d9c4224408b621fb3d3
+size 1717337

anime-style-gurren-lagann-lora-ltx-2-3/example_images/04_video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a232bafad9427418579e173d93ffd9fa3fce1a74a8948c65379bba495379b45d
+size 969681

anime-style-gurren-lagann-lora-ltx-2-3/example_images/05_video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:719ce545db51f06d21e92a86e3c5446e2b3969676eebe81136158f193f8c36ef
+size 1958030

anime-style-gurren-lagann-lora-ltx-2-3/example_images/06_video.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:275717a8f463f46c220be03bde69204438eac770ba544387f45d12b74751f97b
+size 1066274

anime-style-gurren-lagann-lora-ltx-2-3/gl-23-step00040000.comfy.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b40bf5705b100d5ae8af65562b48059683556a750e19ca7a4152f3a6ff2314c7
+size 1348029320

anime-style-gurren-lagann-lora-ltx-2-3/info.txt

@@ -0,0 +1,447 @@
+═══════════════════════════════════════════════════════════════════════════════
+  ANIME STYLE / GURREN LAGANN LORA - LTX 2.3
+═══════════════════════════════════════════════════════════════════════════════
+
+───────────────────────────────────────────────────────────────────────────────
+  IDENTITY
+───────────────────────────────────────────────────────────────────────────────
+  Base model        LTX-2.3
+  Generation mode   Text-to-Video (T2V), Image-to-Video (I2V)
+  Checkpoint        ltx-2.3-22b-dev (full / base)
+  Category          Visual style
+  Author            tazmannner379
+  Source            https://civitai.com/models/2537530
+  Provenance        CivitAI · model 2537530 · version 2851819
+
+───────────────────────────────────────────────────────────────────────────────
+  WEIGHTS
+───────────────────────────────────────────────────────────────────────────────
+  File              gl-23-step00040000.comfy.safetensors
+  Version           V1 40K steps (recomended)
+  Size              1,285.6 MB
+  Format            SafeTensor / unknown precision
+  SHA256            b40bf5705b100d5ae8af65562b48059683556a750e19ca7a4152f3a6ff2314c7
+  Published         2026-04-12
+
+───────────────────────────────────────────────────────────────────────────────
+  TRIGGER WORDS
+───────────────────────────────────────────────────────────────────────────────
+  › "TTGL"
+
+───────────────────────────────────────────────────────────────────────────────
+  RECOMMENDED SETTINGS
+───────────────────────────────────────────────────────────────────────────────
+  Taken verbatim from the author's description:
+
+  Resolution          832x480
+  Frame count         121
+
+───────────────────────────────────────────────────────────────────────────────
+  ADOPTION & POPULARITY
+───────────────────────────────────────────────────────────────────────────────
+  Downloads               1,082
+  Rating                  100% positive (62/62)
+  Comments                0
+  Published               145 days ago
+  Download rate           6.3/day since release
+
+  Adoption rank           #45 of 150 in this collection  (Moderate)
+  Momentum rank           #51 of 150 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 LTX-2.3 ones, not anything specific to this LoRA.
+
+  General LTX-2.3 rules that apply here (see REFERENCE.txt):
+
+  • Stacking an acceleration LoRA with a task LoRA is officially
+    sanctioned: the ID-LoRA template chains LoraLoaderModelOnly(ltx_2.3_22b
+    _distilled_1.1_lora_dynamic_fro09_avg_rank_111_bf16, 0.5) into
+    LoraLoaderModelOnly(ltx-2.3-id-lora-talkvid-3k, 1.0) on the dev
+    checkpoint.
+  • The Python CLI exposes `--lora <path> [strength]` as a repeatable flag
+    with default strength 1.0, so multiple adapters at once is a supported
+    configuration, not a hack.
+  • Load the acceleration/distilled LoRA on the dev checkpoint only. The
+    two-stage pipelines require it for the full model but explicitly do not
+    use it for DistilledPipeline, ICLoraPipeline or DubItPipeline, which
+    already run distilled weights.
+  • Do not stack two distillation LoRAs (e.g. the official rank-384 /
+    rank-111 one plus a community DMD LoRA). The DaSiWa DMD LoRA already
+    blends 75-80% of the official distilled delta into its audio-output
+    route and is described by its author as recreating the distilled model
+    on top of stock dev, i.e. a replacement.
+  • IC-LoRAs are not chained like ordinary LoRAs: each carries a
+    reference_downscale_factor in its safetensors metadata and needs a node
+    that reads it (GetICLoRAParameters -> LTXVAddGuide in core, or
+    LTXICLoRALoaderModelOnly + LTXAddVideoICLoRAGuide in ComfyUI-LTXVideo).
+    Without one the factor silently defaults to 1.
+  • Two IC-LoRA reference guides can be driven from a single IC-LoRA: the
+    CrossView Warp author wires both a depth-warp video and the original
+    video into separate reference guides and sets latent_downscale_factor =
+    1 on both.
+  • DubItPipeline accepts exactly one Dub-It IC-LoRA (`--lora` exactly
+    once) and applies it in both stages — that path is single-adapter by
+    design.
+  • Community convention on strengths when stacking: the IC-LoRA Dual-
+    Character author recommends 0.6-1.0 for a content LoRA used alone,
+    dropped to 0.3-0.5 when combined with others.
+  • Some LoRA pairs are designed to be blended against each other — the LTX
+    2.3 Crisp Enhance / Soft Enhance pair is described by its author as
+    "can be used together and balanced at different strengths".
+  • The gemma-3-12b-it-abliterated LoRA is not a video LoRA and does not
+    belong in the DiT chain — it patches the Gemma text encoder through a
+    model+clip LoraLoader, as the official templates wire it.
+  • The full, flexible, trainable bf16 model — ltx-2.3-22b-dev.safetensors
+    is 46.15 GB bf16, 29.15 GB as ltx-2.3-22b-dev-fp8.safetensors. This is
+    what LoRAs are trained against, and what four of the six official
+    ComfyUI LTX-2.3 templates load (t2v, i2v, ia2v, id_lora). Needs either
+    real guidance (Lightricks' own single-stage workflow sets LTXVScheduler
+    to 15 steps with MultimodalGuider) or the distill
+
+───────────────────────────────────────────────────────────────────────────────
+  PROS
+───────────────────────────────────────────────────────────────────────────────
+  + Well established — 1,082 downloads. [source: CivitAI stats]
+  + Positively received — 62 thumbs-up with no down-votes. [source: CivitAI
+    stats]
+  + Explicit trigger word(s) — TTGL — so the effect can be turned on and
+    off from the prompt. [source: model metadata]
+  + Author documents use across several modes (T2V, I2V). [source:
+    description]
+  + Author published concrete recommended settings. [source: description]
+
+───────────────────────────────────────────────────────────────────────────────
+  CONS & CAVEATS
+───────────────────────────────────────────────────────────────────────────────
+  - Large at 1286 MB — slower to load and heavier to stack. [source: file
+    size]
+
+───────────────────────────────────────────────────────────────────────────────
+  COMPATIBILITY WARNING
+───────────────────────────────────────────────────────────────────────────────
+  Within LTX-2.3, dev vs distilled is low risk: they are the same
+  architecture and the distillation itself is published as a LoRA that the
+  model card calls "applicable to the full model", so ordinary content LoRAs
+  load on both. Expect style/strength drift rather than breakage; the
+  distilled LoRA is meant to be applied to a dev checkpoint, not stacked on
+  an already-distilled one (the Python pipelines that start from a distilled
+  checkpoint — DistilledPipeline, ICLoraPipeline, DubItPipeline — explicitly
+  do not take --distilled-lora). The genuine risk is cross-version and cross-
+  format, and it is silent rather than a hard error: the LTX-2 repo README
+  states flatly that files "are not interchangeable between the two models,
+  and a LoRA only works with the model it was trained on", and a wrong-
+  version LoRA surfaces only as a stream of "lora key not loaded:
+  diffusion_model...." console warnings while generation proceeds as if no
+  LoRA were attached. The same silent no-op hits correctly-versioned LoRAs
+  saved in a non-ComfyUI key layout: SimpleTuner issue #2349 shows keys like
+  diffusion_model.connectors.audio_connector.transformer_blocks.N.attn1.to_q
+  going unloaded, where the LTX-2 trainer's own layout is attn1/attn2 for
+  video, audio_attn1/audio_attn2 for audio and audio_to_video_attn /
+  video_to_audio_attn for cross-modal. Community finetunes (Sulphur 2, JoyAI-
+  Echo merges) load stock LTX-2.3 LoRAs but shift the result. And an IC-LoRA
+  loaded without a metadata-reading node loses its reference scale: ComfyUI's
+  GetICLoRAParameters reads reference_downscale_factor out of the LoRA's
+  safetensors metadata and silently falls back to 1 when it is absent, so a
+  ref0.5 IC-LoRA wired straight into LTXVAddGuide without it feeds the
+  reference at the wrong latent scale.
+
+───────────────────────────────────────────────────────────────────────────────
+  AUTHOR'S DESCRIPTION
+───────────────────────────────────────────────────────────────────────────────
+  Introduction Trigger word: TTGL (stick it at the start of the prompt). Use
+  character tags to generate the characters looks/voices (optional).
+  If the style doesn't trigger just stick the word "anime" at the end of the
+  prompt. But note that it makes the style change a lot, so I recommend it
+  only as a last resort. Also try different seeds.
+
+  This is trained for t2v, all samples are such. I haven't tested with i2v so
+  give it a try and let me know.
+  What is this lora? This lora's purpose is to recreate the style from the
+  anime Tengen Toppa Gurren Lagann and all the main characters in the show.
+  Or you can just use it for anime style without using the characters. This
+  lora does not only give you the style from the anime, but it's trained on
+  every character from the first season (pre-time skip). I put the trigger
+  words/descriptions at the end of this description. If you put them in the
+  prompt they will trigger and come out (some better than others). The
+  mecha's are not really trained well, you're not gonna be able to prompt
+  them in well from my tests. But it can give other kinds of new original
+  mechas. All the character's voices are also trained (japanese dub). To
+  prompt it say for example : char_yoko says in Japanese "sore wa nan daro?!"
+  (You can also just use english if you want but it wont match voice).
+
+  I wanted to see if I could LTX 2.3 to actually do anime style. Like not
+  just looks but motion. And for the most part I think I was able to do it
+  finally. It's a little bit over trained and some characters (nia and yoko)
+  bleed into the base a bit too much, but you can get around it by describing
+  in the prompt those parts. Just understand you often get yoko with red hair
+  and nia with blonde hair with women.
+
+  Recommended Settings I highly recommend you use 2-pass with distill on 0.5
+  str and use 50FPS. Otherwise you will get motion distortion. You still get
+  it with these settings but reduced by a lot more. Anime just doesn't play
+  with LTX. I hope they train the base model on lots of anime in the next
+  version. Use my example workflow for reference if you wanna recreate using
+  my same settings. If anyone has better settings for anime in LTX let me
+  know. Also this style lora will work with other loras but maybe reduce
+  strength on those and keep this on 1 strength.
+
+  Training My old wan gurren lagann dataset was pretty much useless for LTX.
+  It was 16fps and the clips were too short. And it relied on image dataset
+  to teach style. I did an initial test training with it when LTX 2.0 first
+  came out and the result was not great. This time I completely redid from 0
+  this dataset. I used pyscene to split the first season and the music
+  videos. I set the max for each scene 121 frames. More than 70% of the
+  dataset is around 121 frames each, then I have a bunch of 30-90ish frame
+  clips. The long clips are there to teach voice and motion, and the short
+  are more for style and character likeness. It was trained using my
+  captioning tool with gemini. The captioning part was a bit annoying because
+  google has added limits for paying customers. Openrouter doesn't have the
+  limits, but it costs more. I just spent more on openrouter this time and
+  used their api with my caption tool with gemini 3.1 preview since it can
+  understand video + sound (and it does whole videos not just set frames from
+  the video). I am interested to test gemma4 next time.
+
+  The dataset was a bit heavy on Yoko and Nia. If I wanted to do this as a
+  proper style lora, I would reduce their data by a lot. I think next I will
+  try a generic anime style lora with lots of different characters from
+  various shows. I will include some sample data and my config files in a zip
+  in the "dataset" downloads for reference.
+
+  I trained on 832x480 resolution with target fps 24.0. And I padded/trimmed
+  frames which didn't fit into LTX's frame bucket size. (gemini can create a
+  script with FFMPEG to do this for you easy). I learned from my dispatch
+  lora to set the LR on the sound to around half of the LR of the video. This
+  allowed the voices to not over train while the video took its time to
+  catchup. Also its rank 64 (thus the 1.2gb file size) since there is a lot
+  to train for.
+
+  By the way, you can train style very easy on images only. I tried from my
+  wan image dataset and it came out sort of ok. But it didn't have proper
+  anime motion or the look of the show. But I think it has viability in areas
+  outside of anime. 3D, or stopmotion claymotion etc. can be done super quick
+  with image only since I think the base model knows it well enough. The base
+  model doesn't know anime enough.
+
+  This trained locally to around 45K steps, but I personally like the 40K
+  steps checkpoint too. So I add both here and you can decide which is
+  better. I also added a 34K steps version which I think has less over train
+  issues so it's more stable. I think the 40K doesn't bleed yoko/nia into the
+  generations as much and recommend this the most, but the 45K one seems
+  better for anime motion? It took around 80 hours or so to train this far at
+  around 5 s/it locally on a 5090. It was going at 4.5 s/it for a while but
+  it would sometimes OOM (I guess due to the wattage limits I was setting
+  later on) so I put it to 4 block swap and it was trained on float8. I used
+  akanetendo25's musubi fork to train this. Big shout out to him and his
+  excellent implementation of LTX 2.3.
+  Closing thoughts I think this was a good chance to learn about how much we
+  can teach LTX 2.3. And I think the base model will need more training on
+  animation and anime (3D / games seems fine). I know some animation can be
+  triggered in base model but I think rick and morty and spongebob etc. are
+  not enough. They need to rip through a ton of anime (they should try...).
+
+  For the lora's final output, visually its still not as good as Wan 2.1 or
+  2.2. But it kind feels like somewhere in the middle of wan 5b and wan 2.2.
+  Sometimes you have some really bad interaction but one little golden nugget
+  of great animation and anatomy comes out. And there is a feeling that at
+  some point in the future we can get there consistently. I don't wanna give
+  up on trying to get anime or animation in LTX working, so I hope we can get
+  closer with each attempt. And I know its very easy for this lora to go off
+  the rails. Some hands are bad, sometimes doesn't trigger style, or it just
+  stays motionless for whatever reason. But I think after a few tries and
+  adjustments on prompts/seeds we can get something nice.
+
+  At the end of the day, this lora isn't going to teach LTX anatomy or fast
+  motion. Try to make prompts that play into the strengths of what the base
+  model can do. Talking scenes especially are great. And the anime has lots
+  of extreme close ups of people talking actually when you take it apart
+  scene by scene.
+
+  Character triggers Note: Take these with a grain of a salt. I fed the
+  captions into google ai studio and it got really lazy. And I couldn't coax
+  it to properly do it IMO, the latest gemini models suck ass. The order is
+  from most appearing in the dataset to less. Anyone with less than 5
+  appearances is not listed below but may be in the dataset.
+  Character Tag: char_yoko Primary Outfit: A black bikini top with a red
+  flame pattern, black shorts with a white studded belt, a thick white and
+  pink striped scarf around her neck, black fingerless gloves or detached arm
+  warmers with yellow studs, and pink thigh-high stockings with boots. She
+  frequently carries a large sniper rifle. Appearance: A young woman with
+  long, bright red hair tied in a high ponytail, adorned with a white skull
+  hair clip and yellow hair sticks. She has fair skin, golden-yellow eyes,
+  and occasionally wears yellow-tinted goggles resting on her forehead.
+  Character Tag: char_simon Primary Outfit: A blue jacket with a red-lined
+  collar and red accents on the shoulders, a white undershirt, a red armband
+  tied around his left bicep, red goggles resting on his forehead, and a
+  drill necklace. Appearance: A boy with messy dark blue hair, wide eyes, and
+  dirt smudges on his cheeks.
+  Character Tag: char_nia Primary Outfit: A pink and white dress with a
+  yellow collar adorned with two red spheres, and a large pink flower-shaped
+  hair accessory. Appearance: A girl with pale skin, large blue eyes with
+  distinct pink, four-pointed star-shaped pupils, and fluffy pale blonde hair
+  with light blue undersides.
+  Character Tag: char_kamina Primary Outfit: Large, angular, translucent red
+  sunglasses, a tattered red cape with a high collar, white bandages wrapped
+  around his abdomen and forearms, and dark pants with a white studded belt.
+  Appearance: A man with spiky blue hair, purple eyes, and blue tattoos on
+  his upper arms and chest.
+  Character Tag: char_kiyoh Primary Outfit: An olive-green string bikini.
+  Appearance: A woman with long blonde hair and blue eyes.
+  Character Tag: char_leeron Primary Outfit: A high-collared purple and pink
+  jacket with gold studs and round glasses. Appearance: A man with short
+  green hair, purple eyeshadow, pink lipstick, and a beauty mark.
+  Character Tag: char_kittan Primary Outfit: A dark, sleeveless vest
+  revealing his neck and upper chest, dark pants, and a yellow fingerless
+  glove. Appearance: A man with spiky blonde hair and thick prominent
+  eyebrows.
+  Character Tag: char_kinon Primary Outfit: A green top over a black garment,
+  with a large white bow tied at her waist. Appearance: A girl with short
+  brown hair and round glasses.
+  Character Tag: char_kiyal Primary Outfit: A black sleeveless top, black
+  shorts, black thigh-high socks, a white belt, and yellow fingerless gloves.
+  Appearance: A girl with dark purple twin tails and green eyes.
+  Character Tag: char_rossiu Primary Outfit: A white high-collared shirt with
+  green sleeves and purple shoulder stripes. Appearance: A man with a brown
+  ponytail and pale skin.
+  Character Tag: char_gimmy Primary Outfit: A pale green tank top.
+  Appearance: A boy with short spiky red hair and freckles.
+  Character Tag: char_dayakka Primary Outfit: A white tank top, a brown belt,
+  opaque blue goggles, and a thick necklace made of large, shiny, metallic
+  spherical beads. Appearance: A muscular man with dark skin and short green
+  hair.
+  Character Tag: char_lordgenome Primary Outfit: A dark garment with a high,
+  pointed collar. Appearance: A massive, muscular, bare-chested man with dark
+  skin, a thick black beard, and a bald head featuring a dark, vertical
+  marking down the center.
+  Character Tag: char_viral Primary Outfit: A dark reddish-brown high-
+  collared jacket with a red central zipper and red rectangular shoulder
+  pads. Appearance: A man with pale skin, blonde hair that covers his right
+  eye, and sharp, jagged teeth.
+  Character Tag: char_darry Primary Outfit: A yellow top with a white
+  backpack. Appearance: A girl with pink hair.
+  Character Tag: char_adiane Primary Outfit: A magenta top with a black lace-
+  up corset, a red choker, and a white ruffled collar. Appearance: A woman
+  with long dark blue hair, a black right eyepatch, red lipstick, and a
+  thick, segmented brown scorpion-like tail.
+  Character Tag: char_boota Primary Outfit: Sunglasses. Appearance: A small
+  brown pig-mole.
+  Character Tag: char_gurren_lagann Primary Outfit: A face on its torso with
+  yellow V-shaped sunglasses and sharp teeth, topped by a black head with a
+  golden crescent moon crest on its helmet. Appearance: A large red, black,
+  and yellow mechanical robot.
+  Character Tag: char_kidd Primary Outfit: A black sleeveless shirt with a
+  high, flared collar, tan fingerless gloves, and a brown belt. Appearance: A
+  young man with spiky blonde hair featuring a green patch on the side.
+  Character Tag: char_zorthy Primary Outfit: A pink padded vest with dark
+  circles over a blue long-sleeve shirt, and blue sunglasses. Appearance: A
+  man with messy green hair.
+  Character Tag: char_iraak Primary Outfit: A yellow tank top and olive green
+  pants. Appearance: A man with orange spiky hair.
+  Character Tag: char_jorgun Primary Outfit: A dark green vest with metal
+  studs and dark green armbands, and blue star-shaped glasses. Appearance: A
+  man with a brown afro and a beard.
+  Character Tag: char_balinbow Primary Outfit: A dark green vest with metal
+  studs and dark green armbands, and pink star-shaped glasses. Appearance: A
+  man with a brown afro and a beard.
+  Character Tag: char_makko Primary Outfit: A pink striped shirt and a green
+  necklace. Appearance: A man with glasses.
+  Character Tag: char_guame Primary Outfit: A yellow pipe. Appearance: A
+  stout, green reptilian creature with an orange chest.
+  Character Tag: char_enki Primary Outfit: A yellow crest, red eyes, and a
+  black faceplate. Appearance: A tall, humanoid mecha with a white upper
+  body, grey arms and legs, and dark grey accents around its waist and
+  thighs.
+  Character Tag: char_lazengann Primary Outfit: Red circular accents.
+  Appearance: A black mecha with glowing orange circular back ports, a horned
+  head, and a pointed tail.
+  Character Tag: char_cytomander Primary Outfit: A white and teal outfit
+  featuring a high yellow collar with green circles, large white feathery
+  lapels, a gold choker, and teal sleeves. Appearance: A pale man with purple
+  markings under his narrow eyes and white hair styled with a tall red crest
+  on top.
+  Character Tag: char_king_kittan Primary Outfit: Red rectangular side blocks
+  with small arms. Appearance: A golden-yellow metallic surface, a central
+  crest, glowing eyes, and an angular face.
+
+───────────────────────────────────────────────────────────────────────────────
+  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_video.mp4  (video, 2560x1408)
+      still: 01_poster.jpg
+  02_video.mp4  (video, 1920x1024)
+      still: 02_poster.jpg
+  03_video.mp4  (video, 1920x1024)
+      still: 03_poster.jpg
+  04_video.mp4  (video, 1920x1024)
+      still: 04_poster.jpg
+  05_video.mp4  (video, 2560x1408)
+      still: 05_poster.jpg
+  06_video.mp4  (video, 1920x1024)
+      still: 06_poster.jpg
+
+───────────────────────────────────────────────────────────────────────────────
+  HOW TO USE (COMFYUI)
+───────────────────────────────────────────────────────────────────────────────
+  1. Update ComfyUI. LTX-2.3 is supported in ComfyUI core
+  (comfy_extras/nodes_lt.py provides EmptyLTXVLatentVideo, LTXVAddGuide,
+  GetICLoRAParameters, LTXVConditioning, LTXVScheduler,
+  LTXVConcatAVLatent/LTXVSeparateAVLatent, LTXVReferenceAudio and more) and
+  the six official templates need no custom nodes; install the Lightricks
+  "ComfyUI-LTXVideo" custom node pack only if you want IC-LoRA loaders, low-
+  VRAM loaders, HDR decode, audio-only nodes or Q8 nodes.
+  2. Put the base checkpoint in ComfyUI/models/checkpoints/ —
+  ltx-2.3-22b-dev-fp8.safetensors (29.15 GB) or ltx-2.3-22b-distilled-
+  fp8.safetensors (29.53 GB) from Lightricks/LTX-2.3-fp8, or the bf16
+  originals (46.15 GB) from Lightricks/LTX-2.3. It is a single file that also
+  carries the video VAE, the audio VAE and the text projection.
+  3. Put gemma_3_12B_it_fp4_mixed.safetensors in
+  ComfyUI/models/text_encoders/ (from Comfy-Org/ltx-2). The Python package
+  instead wants the full google/gemma-3-12b-it-qat-q4_0-unquantized folder.
+  4. Put ltx-2.3-spatial-upscaler-x2-1.1.safetensors in
+  ComfyUI/models/latent_upscale_models/ (needed by every two-stage template),
+  and moge_2_vitl_normal_fp16.safetensors (Comfy-Org/MoGe) in
+  ComfyUI/models/geometry_estimation/ if you use the IC-LoRA depth path.
+  5. Put every LoRA .safetensors in ComfyUI/models/loras/. Subfolders are
+  fine — Lightricks' own workflows reference
+  "ltxv/ltx2/ltx-2.3-22b-distilled-lora-384-1.1.safetensors".
+  6. Load a normal content/style LoRA with LoraLoaderModelOnly, wired between
+  CheckpointLoaderSimple's MODEL output and the guider/sampler. In the stock
+  templates you chain it after the distilled-LoRA loader. Use the full
+  LoraLoader (model + clip) only for the gemma-3-12b-it-abliterated LoRA,
+  which patches the Gemma text encoder rather than the video model.
+  7. On a dev checkpoint, also load the acceleration LoRA
+  (ltx_2.3_22b_distilled_1.1_lora_dynamic_fro09_avg_rank_111_bf16.safetensors
+  at 0.5) unless you intend to run the slow guided path.
+  8. For an IC-LoRA, the LoRA loader alone is not enough: either use
+  LTXICLoRALoaderModelOnly + LTXAddVideoICLoRAGuide (ComfyUI-LTXVideo), or —
+  as the official ic_lora template does — feed the LoRA-loaded model through
+  core's GetICLoRAParameters into LTXVAddGuide, so the
+  reference_downscale_factor stored in the file's safetensors metadata is
+  read instead of defaulting to 1.
+  9. Template Library > Video > any LTX-2.3 workflow gives you correct wiring
+  to start from; workflow JSONs are at Comfy-Org/workflow_templates/templates
+  /video_ltx2_3_{t2v,i2v,ia2v,flf2v,ic_lora,id_lora}.json and
+  Lightricks/ComfyUI-LTXVideo/example_workflows/2.3/.
+
+───────────────────────────────────────────────────────────────────────────────
+  Scraped by lorakit on 2026-09-04 23:16 UTC.
+  Stats and description are the author's; pros/cons are derived from the
+  evidence tagged beside each line.
+───────────────────────────────────────────────────────────────────────────────

anime-style-gurren-lagann-lora-ltx-2-3/metadata.json

@@ -0,0 +1,146 @@
+{
+  "name": "Anime Style / Gurren Lagann Lora - LTX 2.3",
+  "slug": "anime-style-gurren-lagann-lora-ltx-2-3",
+  "base_model": "LTX-2.3",
+  "media_type": "video",
+  "model_id": 2537530,
+  "version_id": 2851819,
+  "version_name": "V1 40K steps (recomended)",
+  "source": "civitai",
+  "url": "https://civitai.com/models/2537530",
+  "repo_id": null,
+  "file_path": null,
+  "author": "tazmannner379",
+  "category": {
+    "key": "style",
+    "label": "Visual style"
+  },
+  "modalities": [
+    "t2v",
+    "i2v"
+  ],
+  "primary_modality": "t2v",
+  "checkpoint_family": "dev",
+  "trigger_words": [
+    "TTGL"
+  ],
+  "recommended_settings": {
+    "Resolution": "832x480",
+    "Frame count": "121"
+  },
+  "chaining": {
+    "author": {
+      "stacks": "not stated",
+      "notes": "",
+      "avoid": [],
+      "requires": [],
+      "order": "",
+      "strength": "",
+      "stated": false
+    },
+    "general_rules": [
+      "Stacking an acceleration LoRA with a task LoRA is officially sanctioned: the ID-LoRA template chains LoraLoaderModelOnly(ltx_2.3_22b_distilled_1.1_lora_dynamic_fro09_avg_rank_111_bf16, 0.5) into LoraLoaderModelOnly(ltx-2.3-id-lora-talkvid-3k, 1.0) on the dev checkpoint.",
+      "The Python CLI exposes `--lora <path> [strength]` as a repeatable flag with default strength 1.0, so multiple adapters at once is a supported configuration, not a hack.",
+      "Load the acceleration/distilled LoRA on the dev checkpoint only. The two-stage pipelines require it for the full model but explicitly do not use it for DistilledPipeline, ICLoraPipeline or DubItPipeline, which already run distilled weights.",
+      "Do not stack two distillation LoRAs (e.g. the official rank-384 / rank-111 one plus a community DMD LoRA). The DaSiWa DMD LoRA already blends 75-80% of the official distilled delta into its audio-output route and is described by its author as recreating the distilled model on top of stock dev, i.e. a replacement.",
+      "IC-LoRAs are not chained like ordinary LoRAs: each carries a reference_downscale_factor in its safetensors metadata and needs a node that reads it (GetICLoRAParameters -> LTXVAddGuide in core, or LTXICLoRALoaderModelOnly + LTXAddVideoICLoRAGuide in ComfyUI-LTXVideo). Without one the factor silently defaults to 1.",
+      "Two IC-LoRA reference guides can be driven from a single IC-LoRA: the CrossView Warp author wires both a depth-warp video and the original video into separate reference guides and sets latent_downscale_factor = 1 on both.",
+      "DubItPipeline accepts exactly one Dub-It IC-LoRA (`--lora` exactly once) and applies it in both stages — that path is single-adapter by design.",
+      "Community convention on strengths when stacking: the IC-LoRA Dual-Character author recommends 0.6-1.0 for a content LoRA used alone, dropped to 0.3-0.5 when combined with others.",
+      "Some LoRA pairs are designed to be blended against each other — the LTX 2.3 Crisp Enhance / Soft Enhance pair is described by its author as \"can be used together and balanced at different strengths\".",
+      "The gemma-3-12b-it-abliterated LoRA is not a video LoRA and does not belong in the DiT chain — it patches the Gemma text encoder through a model+clip LoraLoader, as the official templates wire it.",
+      "The full, flexible, trainable bf16 model — ltx-2.3-22b-dev.safetensors is 46.15 GB bf16, 29.15 GB as ltx-2.3-22b-dev-fp8.safetensors. This is what LoRAs are trained against, and what four of the six official ComfyUI LTX-2.3 templates load (t2v, i2v, ia2v, id_lora). Needs either real guidance (Lightricks' own single-stage workflow sets LTXVScheduler to 15 steps with MultimodalGuider) or the distill"
+    ]
+  },
+  "pros": [
+    "Well established — 1,082 downloads. [source: CivitAI stats]",
+    "Positively received — 62 thumbs-up with no down-votes. [source: CivitAI stats]",
+    "Explicit trigger word(s) — TTGL — so the effect can be turned on and off from the prompt. [source: model metadata]",
+    "Author documents use across several modes (T2V, I2V). [source: description]",
+    "Author published concrete recommended settings. [source: description]"
+  ],
+  "cons": [
+    "Large at 1286 MB — slower to load and heavier to stack. [source: file size]"
+  ],
+  "file": {
+    "name": "gl-23-step00040000.comfy.safetensors",
+    "size_bytes": 1348029320,
+    "size_mb": 1285.58,
+    "sha256": "b40bf5705b100d5ae8af65562b48059683556a750e19ca7a4152f3a6ff2314c7"
+  },
+  "stats": {
+    "downloads": 1082,
+    "thumbs_up": 62,
+    "thumbs_down": 0
+  },
+  "popularity": {
+    "downloads": 1082,
+    "thumbs_up": 62,
+    "thumbs_down": 0,
+    "comments": 0,
+    "approval": "100% positive (62/62)",
+    "age_days": 145.4,
+    "downloads_per_day": 6.29,
+    "rank_by_downloads": 45,
+    "rank_by_momentum": 51,
+    "of_total": 150,
+    "adoption_tier": "Moderate",
+    "momentum": "Steady"
+  },
+  "selection_score": 2226.0,
+  "selection_reasons": [
+    "1,082 downloads",
+    "62 thumbs-up",
+    "documented triggers + usage notes",
+    "top-ranked for its category"
+  ],
+  "example_media": [
+    {
+      "file": "01_video.mp4",
+      "type": "video",
+      "resolution": "2560x1408",
+      "prompt": "",
+      "poster": "01_poster.jpg"
+    },
+    {
+      "file": "02_video.mp4",
+      "type": "video",
+      "resolution": "1920x1024",
+      "prompt": "",
+      "poster": "02_poster.jpg"
+    },
+    {
+      "file": "03_video.mp4",
+      "type": "video",
+      "resolution": "1920x1024",
+      "prompt": "",
+      "poster": "03_poster.jpg"
+    },
+    {
+      "file": "04_video.mp4",
+      "type": "video",
+      "resolution": "1920x1024",
+      "prompt": "",
+      "poster": "04_poster.jpg"
+    },
+    {
+      "file": "05_video.mp4",
+      "type": "video",
+      "resolution": "2560x1408",
+      "prompt": "",
+      "poster": "05_poster.jpg"
+    },
+    {
+      "file": "06_video.mp4",
+      "type": "video",
+      "resolution": "1920x1024",
+      "prompt": "",
+      "poster": "06_poster.jpg"
+    }
+  ],
+  "tags": [
+    "style"
+  ],
+  "scraped_at": "2026-09-04T23:16:58+00:00",
+  "path": "ltxv-2.3/anime-style-gurren-lagann-lora-ltx-2-3"
+}