Download
1 variant available
678
3.5K
Generation, training and LoRA distribution on Civitai are covered by Civitai’s own license agreement with MiniMax. If you download these weights and run them yourself, your use is instead governed by the MiniMax H3 Community License Agreement, whose grant excludes the European Union, the United Kingdom, the Republic of Korea and the United States of America.
MiniMax H3
MiniMax H3 general video generation workflow (VDN 12-step HD edition). Generate high-definition video with text-native audio from keyframes and reference images, without any external audio input.
Main features:
- Inputs: first frame, last frame, up to 3 reference images, plus an optional source video used for frame reference only (its audio is not used)
- Official Duration Planner for text-native audio (17n+5 frame grid)
- OpenVDN DMD8 two-pass pipeline: 8NFE base pass, learned latent x2 upscale to 1920x1088 / 1088x1920 (16:9 or 9:16), 4NFE refine (12-step total)
- Audio audit, AV decode, output trim and safe AV save at approx 1080P / 24fps
Suggested workflow:
1. Provide first/last frames and reference images with the same aspect ratio
2. Optionally load a source video as visual reference only
3. Run the graph; the planner, both passes, upscale and audio audit chain runs automatically, then export the approx 1080P save
Dependencies:
- minimax_h3_fl_ref_hybrid_scheme_a_25-49_bf16.safetensors (H3 ref UNET)
- qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors (CLIP)
- minimax_h3_video_vae_fp16.safetensors and minimax_h3_audio_vae_fp32.safetensors
- minimax_h3_latent_upscaler_3d_fp16.safetensors (learned latent x2)
- OpenVDN/vdn-minimax-h3 (DMD8 stage), KSK T8 custom nodes
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2108580251193323522?inviteCode=rh-v1111
Bilibili tutorial video: https://www.bilibili.com/video/BV1pTph6EEUv/
RunningHub registration with invite code rh-v1111 (1000 RH coins): https://www.runninghub.ai/?inviteCode=rh-v1111