RDBT | Anima

v1.0: I think it's time to release v1.0. I've migrated to the krea2 model (never imagined that krea2 would be my next choice...). Fine-tuning the anima model is really fun.

I also uploaded int8 version for ComfyUI hardware int8 (labeled as fp8).

Dataset doesn't change much. But distilled as 8-step model. If you want higher diversity, you can try v0.39.b.

RDBT [Anima]

This is a finetuned model with 10k high aesthetic images paired with natural language captions from LLM. Then distilled to further improve quality and stability. Due to personal needs, training data does not contain any shiny plastic glossy AI image.

It's not overfitted and doesn't have a default style. (Might sound strange for some people, see FAQ.)

I use it as a clean starting point to stack more style LoRAs. I can stack whatever I want and get exactly what I stacked.

See this page for update log.

For advanced users: The RDBT model is trained as LoRA natively. See this page for original LoRA, update more frequently.

This model is based on:

• prefix with ym: AnimaYume (hf link) (civitai link).

• prefix with b,p: Anima pretrained (hf link)

Sharing merges using this model is not allowed. If someone is selling this model as their own, I'm happy to list them here so everyone knows.

Known model thieves: NukeA.I (behind paywall on tensorart).

I wrote a story about it. Also contains a guide for trainers about "how to bake special trigger word into your model".

Usage:

Settings:

CFG scale: 1~3. This model has been guidance distilled. You can disable CFG (CFG 1) and run the model 2x faster. Cover images are without CFG for demonstration.

Steps: 16+

Prompt

Always specify style, or use a style LoRA. Otherwise, you will get random/mixed style.

This is a feature, not a bug. This model does not provide a overfitted default style.

Quality tags:

It's recommended to omit all the quality tags, or just keep the "masterpiece", if you're not confident. Omitting those redundant tokens allows LLM to pay more attention on other words.

Quality tags have been reinforced during distillation. Thus they don't have noticeable effects. Same as negative tags. If you use cfg, there is no need to dump "score_1, blurry, worst quality, jpeg artifacts, extra arms,... x100 words" into your negative prompt. Those things have been distilled out.

FAQ:

What's wrong with the "default style"?

If an image generation model has a default style, it means the model will unconditionally generate that style, even if you don't specify it in the prompt. This phenomenon has a technical name: overfitting.

If you like this default style, then great, it's always here and you even don't have to prompt. But if you want to generate something other than the default style, unfortunately, you'll find that you cannot overwrite it.

FYI: 95% ckpts you see out there, are just base ckpt + overfitted style LoRAs.

Training settings

All captions are NL from Google Gemini.

Optimizer: adamw, constant lr 0.00002, weight decay 0.1, batch size 16.

LoRA rank/alpha 24.

Timesteps shift 3.

Block 0-2 and adaln linear layers are skipped.