JusteLeo/Anima-GGUF

Anima - GGUF Quantized Models

These are GGUF quantized versions of the Anima model, optimized for local inference on low-RAM devices using stable-diffusion.cpp.

Available Formats

• anima-preview-Q4_K.gguf : Smallest footprint. Best for Low RAM.
• anima-preview-Q5_K.gguf : Balanced size and quality.
• anima-preview-Q6_K.gguf : High quality, requires slightly more memory.

How to Use

Ensure you are using the latest version of stable-diffusion.cpp (Anima support is integrated into the main branch).

./bin/sd-cli \
    --diffusion-model /path/to/anima-preview-Q4_K.gguf \
    --llm /path/to/qwen_3_06b_base.safetensors \
    --vae /path/to/qwen_image_vae.safetensors \
    --vae-tiling \
    --steps 30 \
    --cfg-scale 4 \
    -W 1024 -H 1024 \
    -p "a cute blue cat in a park" \
    -o output.png

(Note: The --vae-tiling flag is recommended to prevent out-of-memory errors during the image decoding phase).

Acknowledgments and Credits

• Original Model: circlestone-labs.
• Engine: The stable-diffusion.cpp project for making local AI accessible.
• Implementation: Special thanks to GitHub contributor @rmatif for his work on the implementation of Anima support.
• License: These files are subject to the circlestone-labs-non-commercial-license.