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.

Some cover images look ridiculous; they're just for demonstration purposes for corner cases.

I use this model as a clean starting point to stack more LoRAs. The dataset is big enough, model was not overfitted. There is no default style (overfitted style, which ignores prompt and is always active) that conflicts with your style LoRA. You can get the style exactly as it should be.

Update log and version info: link.

For those who don't want to stack style LoRAs and is looking for a out-of-the-box ckpt with high aesthetic style: RDBT | Anime, which is based on RDBT finetuned base, but optimized for high quality digital art.

For advanced users: RDBT model is trained as LoRA natively. You don't need to download this full ckpt. Original LoRA: link.

Links to base model:

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

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

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+

• Sampler: Euler (best diversity), Euler a/er_sde etc. (better stability)

• Res: 1MP

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.

My homemade distillation model is not a "turbo" model.

The unwritten definition of "turbo" model is 8-step + aesthetic RL + many black magic. My distillation is just a basic distillation LoRA that mimics what the original model would generate. A real "turbo" model is too much for me.

However, good things: Compared to the official turbo model, this distillation has very high diversity (seed, style diversity, etc. Because it needs 16 steps), and is more compatible with style LoRAs (it won't add a quality filter which changes/overwrites the style drastically).

Sharing merges using this model is not allowed.

This mode is a free and it will always be free. This "restriction" won't affect anyone. It's only aimed at those who steal others' models to sell.

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".