morpfvr94/minimax-h3-worker-weights

MiniMax H3 worker weights

Unmodified copies of the files one ComfyUI reference-to-video worker loads, gathered in one small repository so a serverless host can cache only what it needs. No file has been changed.

The owner of this repository did not create any of these files and claims no authorship, ownership or credit. See CREDITS.md for the authors and sources. For new versions and support, go to those sources.

Terms

MiniMax H3 is licensed under the MiniMax H3 Community License Agreement, Copyright © 2026 MiniMax. All Rights Reserved. The full agreement is in LICENSE.

By downloading or using the MiniMax H3 files here you accept that agreement directly, including its territorial scope, the use restrictions in Section V and the Acceptable Use Policy in Exhibit A. Those restrictions apply to you. Other files keep the terms of their source, as listed in CREDITS.md.

Layout

File	Size	Purpose
diffusion_models/minimax_h3_ref2va_pruned_int8_convrot.safetensors	20.97 GB	Reference-to-video model, int8
text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors	15.69 GB	Text encoder
vae/minimax_h3_video_vae_int8_convrot.safetensors	2.81 GB	Video VAE
vae/minimax_h3_audio_vae_fp32.safetensors	0.61 GB	Audio VAE
loras/minimax_h3_ref2v_turbo_8step_v1.1_768p_comfyui_bf16.safetensors	1.96 GB	8-step turbo LoRA for ref2va
loras/minimax_h3_ref2v_turbo_4step_v1.0_768p_comfyui_bf16.safetensors	1.96 GB	4-step turbo LoRA for ref2va
loras/minimax_h3_fl2v_turbo_4step_v1.3_768p_comfyui_bf16.safetensors	1.96 GB	4-step turbo LoRA for fl2va, a model not stored here
loras/h3-realism-people-t2v-i2v-r2v.safetensors	0.13 GB	People realism LoRA

The folders match ComfyUI's model folders, so this repository can be used as an extra_model_paths base.