One ComfyUI node replaces 42-step MiniMax H3 ref-to-video workflow
Dudelydad78 · reddit · 2026-10-23
Reddit user IronLeo78 released [WIP] Minimax Power Producer, a single ComfyUI node that folds the entire MiniMax H3 ref-to-video pipeline—previously 42 node-stages—into one click-to-run node. Built with Claude's help and tested on i9-14900K / RTX 5080 / 64GB.
Key features
- Up to 3 reference images, 1 reference video, plus custom audio, with labeled thumbnails
- i2v and v2v; long clips processed segment by segment ("Next Segment On Run")
- Audio trim, track select, and per-segment chunking; can use a video's audio only without decoding frames; MiniMax-generated audio or bypass
- Latent Upscale toggle (2x/3x/4x): off for fast drafts, on for finals, same seed saved as upscaled for easy A/B
- Fixed 24fps and known-good sampler settings; STATUS output maps <Picture i>/<Video 1>/<Audio j> tags to files
Setup: update ComfyUI (includes MiniMax H3 nodes), drop into customnodes, place models, restart. Node zip, upscaler, docs and troubleshooting guide on the Hugging Face bucket.
More from coding & agent
- Selfdrive AI agent turns an App Store link into a full user-acquisition plan — PrajwalTomar_ · 2026-10-23
- Thorsten Ball mocks SaaS firms building internal agent platforms nobody uses — HankYeomans · 2026-10-23
- Voice bot team: GPT Live 1 and Gemini 3.8 Thinking both fail in production — Mr_Nice_ · 2026-10-23
- KURE rerankers demoed: 149M nano model boosts the right answer to top spot — tomaarsen · 2026-10-23
- Google launches Gemini API CLI preview: agents generate media with one command — rseroter · 2026-10-23
- Lucas Meijer's litmus test: hard tasks are ones agents fail at when you just ask — lucasmeijer · 2026-10-23