Frozen ComfyUI model pack for:
GitHub cannot host ~91 GB of weights; this HF repo is the one-shot weight mirror.
ComfyUI/models/)diffusion_models/
minimax_h3_fl2va_pruned_int8_convrot.safetensors
minimax_h3_ref2va_pruned_int8_convrot.safetensors
text_encoders/
qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors
H3/qwen3vl_32b_h3_ultra_uncensored_heretic_int8_convrot.safetensors
H3/qwen3vl_32b_h3_generation_tail_50_63_int8_convrot.safetensors
vae/
minimax_h3_video_vae_fp16.safetensors
minimax_h3_audio_vae_fp32.safetensors
upscale_models/
RealESRGAN_x2plus.pth
RealESRGAN_x4plus.pth
hf download drowzeys/keys-heretic-MiniMax-H3-sol-engine-more-DGX-Spark-weights \
--local-dir ~/comfy/ComfyUI/models
Or use the GitHub repo ./install.sh which prefers this pack.
Redistribution of third-party weights is subject to their original licenses. This pack is a convenience mirror for the keys-heretic Spark recipe.
text_encoders/H3/qwen3vl_32b_h3_ultra_uncensored_heretic_int8_convrot.safetensors 26 GBdiffusion_models/minimax_h3_fl2va_pruned_int8_convrot.safetensors 21 GBdiffusion_models/minimax_h3_ref2va_pruned_int8_convrot.safetensors 21 GBtext_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors 16 GBtext_encoders/H3/qwen3vl_32b_h3_generation_tail_50_63_int8_convrot.safetensors 7.6 GBvae/minimax_h3_video_vae_fp16.safetensors 5.2 GBvae/minimax_h3_audio_vae_fp32.safetensors 605 MBupscale_models/RealESRGAN_x2plus.pth 67 MBupscale_models/RealESRGAN_x4plus.pth 67 MB有文件?提交种子。
提交种子Comfy-Org/MiniMax-H3