import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("circlestone-labs/Anima", dtype=torch.bfloat16, device_map="cuda")
pipe.load_lora_weights("leoftm/shimabara_yuuhi_neo_ranga_anima-preview3_lora")
prompt = "[Base:] masterpiece, best quality, highres, absurdres, newest, score_8, score_9, [Character:] 3girls and 1boy, (mid teen girl-shimabara ushio-red short hair-purple eyes-fair skin), (tall mature female-shimabara minami-brown long hair-brown eyes), She is the tallest. (petite-loli-shimabara yuuhi-black long hair-brown eyes-fair skin-small breasts), (youngest boy-joel \\(neo ranga\\)-green hair with single braid-blue eyes-shota ), He is the shortest. [Pose:] The three girls and one boy are standing closely together in a relaxed group formation. (screen left-shimabara ushio casually puts her arm around shimabara minami waist and smiles-their sides pressing firmly together). (center left-shimabara minami gently rests her left hand on center right girl shoulder). (center right-shimabara yuuhi stands slightly in the foreground-turning back to look at the others with a soft smile). (screen right-youngest boy joel \\(neo ranga\\) stands-lightly gripping the strap of his small shoulder bag-leaning in slightly fits naturally into the circle). [Environment_&_Cinematography:] The camera frames them in a perfectly composed medium full shot from a neutral eye-level angle. Bright, natural daylight casts soft, flattering shadows across their figures, highlighting a clear blue sky and the softly blurred green foliage of a serene park environment in the background."
image = pipe(prompt).images[0]shimabara_yuuhi_neo_ranga_anima-preview3_lora_20260414_ep-000020
- Prompt
- [Base:] masterpiece, best quality, highres, absurdres, newest, score_8, score_9, [Character:] 3girls and 1boy, (mid teen girl-shimabara ushio-red short hair-purple eyes-fair skin), (tall mature female-shimabara minami-brown long hair-brown eyes), She is the tallest. (petite-loli-shimabara yuuhi-black long hair-brown eyes-fair skin-small breasts), (youngest boy-joel \(neo ranga\)-green hair with single braid-blue eyes-shota ), He is the shortest. [Pose:] The three girls and one boy are standing closely together in a relaxed group formation. (screen left-shimabara ushio casually puts her arm around shimabara minami waist and smiles-their sides pressing firmly together). (center left-shimabara minami gently rests her left hand on center right girl shoulder). (center right-shimabara yuuhi stands slightly in the foreground-turning back to look at the others with a soft smile). (screen right-youngest boy joel \(neo ranga\) stands-lightly gripping the strap of his small shoulder bag-leaning in slightly fits naturally into the circle). [Environment_&_Cinematography:] The camera frames them in a perfectly composed medium full shot from a neutral eye-level angle. Bright, natural daylight casts soft, flattering shadows across their figures, highlighting a clear blue sky and the softly blurred green foliage of a serene park environment in the background.
- Negative Prompt
- worst quality, low quality, score_1, score_2, score_3, 6 fingers, 6 toes, ai-generated, bad eyes, bad pupils, bad iris, bad hands, bad fingers,
Model Description
Lora hashes: "shimabara_yuuhi_neo_ranga_anima-preview3_lora_20260414_ep-000020: ae322e87aa22"
This is a multi-character LoRA model featuring Yuuhi from Neo Ranga as the main character, designed for anima preview3.
Note: Due to the training data, no mouth outputs appear frequently. This will need to be addressed in future tag updates. For now, mitigate it by adding "no mouth" to the negative prompt.
A LoRA weight around 0.8 is recommended. At 1.0, poses tend to break, while at 0.6, fidelity drops. Adjust up or down depending on your use case.
What "_no_te" ver means: this is a version with the TE trained component removed afterward. Output quality remains the same, and the LoRA file size is smaller, so this version may be sufficient.
Trigger words
You should use shimabara yuuhi to trigger the image generation.
You should use joel \(neo ranga\) to trigger the image generation.
You should use shimabara ushio to trigger the image generation.
You should use shimabara minami to trigger the image generation.
Download model
Download them in the Files & versions tab.
- Downloads last month
- 33
Model tree for leoftm/shimabara_yuuhi_neo_ranga_anima-preview3_lora
Base model
nvidia/Cosmos-Predict2-2B-Text2Image