Rapid-H3
Type: ComfyUICategory: Generation
Rapid-H3 is a ComfyUI workflow or node pack in our Generation category. Apache-2.0 ComfyUI node pack: MiniMax H3 video+audio 3-5x faster on one NVIDIA DGX Spark (about 117 s vs 590-630 s stock for a 5.9 s clip, per the author) via INT8 sm_121 kernels and a 6-step draft-and-refine recipe. H3 weights keep their own licence.
Where it fits: ComfyUI, Open-weight video & audio models.
Listing isn’t endorsement. This is our summary of the project’s own docs. Check the licence and requirements at the source before you build on it.
Lab posts that cover Rapid-H3
Related Lab posts
- Fish Audio Drama 3 final weights top open TTS arenas
- NVIDIA's Open-H3-Regenerate-2K: a three-step 2K pass for MiniMax H3
- Mickmumpitz's one-model H3 short film and Singularity V2
- OPENjutsu: Comfy open-sources a character-swap app on MiniMax H3, and a fully local fork follows within hours
- Henge: an open studio for H3 character swap, motion transfer and head swap, plus Higgsfield open-sources its 'Inside AI' short
Similar skills & repos
- OpenField · Repo, Generation
- Kitty Open Studio · MCP, Generation
- Minimax-h3 Singularity V2 · ComfyUI, Generation
- Henge · ComfyUI, Generation
- OPENjutsu · ComfyUI, Generation
- openjutsu (local fork) · ComfyUI, Generation