RDBT | Anima

RDBT [Anima]

This is a general finetuned + distilled model.

Dataset contains ~10k handpicked images with accurate NL captions from LLM, body/hands anatomy. Does not contain any shiny plastic glossy AI image.

It does not provide any style nor shiny quality filter. This model is for better stability and LoRA compatibility, mitigating LoRA overfitting, and provide the style as-is. Some cover images look ridiculous; they're just for demonstration purposes.

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

See this page for update log and version info.

For advanced users: RDBT model is trained as LoRA natively. See this page for original LoRA.

Base model:

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

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

Sharing merges using this model is not allowed. This "restriction" won't affect anyone. It's only aimed at those who steal others' models to sell. If someone is selling this model as their own, I'm happy to list them here so everyone knows.

Known model thieves: NukeA.I (selling this model 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: 1~3. This model has been distilled. You can disable CFG (CFG 1) and run the model 2x faster. Cover images are without CFG for demonstration. "RenormCFG" node is highly recommended if CFG is enabled (CFG > 1), set "renorm_cfg" value to 1.1.

• Steps: 16+

Prompt:

Always specify style in prompt, or use a style LoRA. Otherwise, you will get random/mixed style. This is a feature, not a bug. This model does NOT have overfitted default style (which ignores prompt and is always active).

Quality tags:

Omit ALL quality tags. You don't need those. The fine-tuning dataset has higher quality than "masterpiece". Thus quality tags don't have effects. Omitting those redundant tokens allows LLM to pay more attention on other words.

Misc.

Distillation:

Homemade 16-step weak distillation based on dmd2. Compared to common dmd2 model, this modified dmd2:

• can generate image very close to what the original model would generate. (e.g.)

• preserved almost all the diversity, every seed is different, just as the original model.

• LoRA friendly.

• targets are sampled with DynamicCFG (mitigated latent overshoot issue in standard CFG, which is the cause of oversaturated/burnt images. Will try to add it in ComfyUI, TBD).