Emanon14/Nameless-Anima

Showcase

Notice: v1.1 Update

Nameless-Anima11.safetensors has been added as version 1.1.

This version was created by rebuilding the Style LoRA used for the model merge. The previous version, Nameless-Anima10.safetensors, will remain available.

Main Changes

• Rebuilt the Style LoRA from scratch.
• Merged the new Style LoRA to create v1.1.
• Improved overall style stability.
• Output may differ slightly from v1.0 in line quality, color tone, texture, and character rendering.

If you prefer the previous output style, please use Nameless-Anima10.safetensors.

Style LoRA Training Details

The Style LoRA used for v1.1 was trained on a dataset of 500 images generated with my custom merged model, NONAMEmix-Vpred_v3.

Training settings:

• Learning rate: 4e-5
• Self-Attention learning rate: 0
• Cross-Attention learning rate: 0
• Rank: 4
• Alpha: 4
• Batch size: 2
• Gradient accumulation steps: 2
• Training steps: 1,000

Model Description

This model was created by merging a custom LoRA into the base model, anima-base-v1.0. The LoRA was trained on a small dataset in order to stabilize the generation style.

Usage & Generation Settings

For details regarding generation settings, prompting guidelines, and the use of various tags, please refer to the original Anima model documentation.

License

The license for this model strictly follows the license of the base model, Anima. Please refer to the original Anima license for exact terms and conditions of use.