🌸 Sakura MiniMax-H3 Uncensored Image-to-Video

Demo scaffold for the uncensored text encoder from webmp3/Sakura-MiniMax-H3-Uncensored-GGUF.

Status: Static Space (free). Full Gradio + ZeroGPU requires a PRO subscription or an account older than 30 days for free ZeroGPU.

This page is a starting point. The full I2V pipeline needs the Sakura GGUF TE + MiniMax-H3 denoiser + VAEs, run via stable-diffusion.cpp or ComfyUI.

What you need for local results

Text encoder (uncensored): files from the Sakura repo (e.g. Q4_K_M)
Denoiser + VAEs: from leejet/MiniMax-H3-GGUF or Comfy-Org
Runtime: stable-diffusion.cpp with MiniMax-H3 support

Example command (local)

sd-cli -M vid_gen \
  --diffusion-model minimax_h3_fl2va-Q4_K_M.gguf \
  --vae minimax_h3_video_vae_fp16.safetensors \
  --audio-vae minimax_h3_audio_vae_fp32.safetensors \
  --llm Sakura-MiniMax-H3-TextEncoder-Uncensored-Q4_K_M.gguf \
  --init-img your_image.png \
  -p "your prompt here"

Adult content only. You are responsible for outputs and applicable laws.

To upgrade this Space to a full Gradio demo with compute, subscribe to Hugging Face PRO and I can push a working app.py.