Note about cover images: Raw output at 1024px. No upscale, no face/hand detailer, no negative prompt. For demonstration purposes only. Workflow included.

See Update Log section for version info.

RDBT [Anima]

Finetuned circlestone-labs/Anima.

ALWAYS experimental, and will be updated randomly as needed.

This model is primarily designed to meet the high aesthetic level that humans can "draw" (aka, 2D illustration), add more details, better textures, and reduce deformed artifacts. It doesn't provide stable style. I use it as a starting point to stack more styles.

About dataset:

• Includes a wide range of images that are "logical and can be accurately described". Common enhancements such as eyes, faces, hands, clothes, lightings, backgrounds, etc.

• Zero AI slop.

• All captions are natural language from Gemini.

This model was trained as a LoRA for better training and distribution efficiency.

Then CFG distilled for better quality and stability. See this img, the sampling process of distilled model is smooth and stable.

Restrictions: Sharing merges using this model is not allowed. If you think this LoRA is useful, please share the link or the LoRA file.

If you are using a "custom" base model and this LoRA "breaks" the base model, instead of blaming the LoRA, try to look at the problem from a different angle: "Has this custom base model already merged this LoRA? And you merged it twice and the weights collapsed." Beware of fake base model creators. Some "creators" never do the training, they only grab other people's models, merge them, wipe all metadata and credits, and sell it as their own base model.

Usage:

Base model

This is a LoRA (a small patch of weights). You will need the base model (full weights) that trained the LoRA.

If you don't know what above means, just want to use a "finetuned base model" without LoRA, you can download this base model, which has merged this LoRA: https://civitai.com/models/2356447.

Prompt

Prefer natural language prompt. Prompt structure: style, subject, action, background.

You can omit all the quality tags. The average output quality of this model should be higher than so-called "masterpiece".

There are two "rough" trigger words:

• "digital anime illustration": 2d anime.

• "digital art", 2d art but not anime. (not many samples)

Recommended settings:

CFG distilled:

• "er_sde", "euler" or "euler_a" sampler.

• 20~30 steps.

• cfg scale 1~3. Note: cfg 1 = disable negative prompt = 2x faster. Cover images are using cfg 1. cfg > 1 gives you better prompt adherence.

• "Adjust Contrast" node to boost contrast.

DMD2 8-steps distilled:

• "euler" or "euler_a" sampler.

• 8~12 steps.

• cfg scale 1.

Advanced (comfyui):

You can mix the sampling process with both models. Let dmd2 process the first 20% to obtain stable high level compositions, then let the cfg distilled or vanilla process the rest to obtain richer details. You can reference this workflow to mix multiple ksamplers: https://civitai.com/images/125707307

Update log

f = finetuned, d = cfg distilled, dmd2 = distribution matching distillation.

Recommended versions (personal opinion):

1. p3 v0.24fd: best aesthetic, 2x faster

2. p2 v0.23f dmd2 b: maximum stability, 8 steps, 8x faster,

===============

Next (planned):

• Improve dmd2 quality.

• 1536px training. (Personally I prefer using other bigger model to upscale, like flux klein 4b.)

(4/8/2026) p3 v0.24fd: Rebased on preview3. -40% steps than v0.23. Less overfitted (?).

(4/4/2026) p2 v0.23f dmd2 b: Different distillation settings.

(4/4/2026) p2 v0.23f dmd2: 8 steps dmd2. First dmd2 anima (?).

(3/28/2026) p2 v0.23fd: Rebased on preview2. Distillation: improved small details and stability (removed a regularization in distillation target and changed to second-order method).

Voting result: v0.20fd won. Thanks for the feedbacks.

(3/24/2026) preview1 v0.20fd b: Distillation: Different settings optimized for anime, high contrast and saturation.

(3/23/2026) p1 v0.20fd: Dataset: More furry. Finetuned base model: from v0.12 + more 100% steps. Distillation: Fixed noisy pixels this time, really.

(3/14/2026) preview v0.19fd b:

Updated dataset. Some private datasets have been dropped. You might notice the style changed.

Fixed high-freq artifacts in v0.12, now you should get a clear image without noisy pixels

b: Testing new distillation settings. Higher contrast. Aligned with common anime art.

(2/19/2026) preview v0.12fd:

Better stability and details, extended dataset.

(3/8/2026) preview v0.11fd 512px:

Prove of concept version for v0.12. Same dataset and settings as v0.12, except it was trained with 512px res.

Released by request, as it might be very useful. Running model in 512px and cfg1 is extremely fast (x10 faster, e.g. 30s -> 3s). If you don't have a beefy GPU. You can use this version to test your ideas/prompts in few seconds.

(2/12/2026) preview v0.6d:

CFG distilled only. No finetuning. Cover images are using Animeyume v0.1.

(2/3/2026) preview v0.2fd:

Speedrun attempt, mainly for testing the training script. Limited training dataset. Only covered "1 person" images plus a little bit of "furry". But it works, and way better than what I expected.