Users Run MiniMax H3 Video Generation Locally on Consumer GPUs

Users report running MiniMax H3 video generation locally on consumer GPUs, from an RTX 3060 12GB via ComfyUI's default ref2va workflow to an optimized RTX 4080 pipeline using INT8 pruning and Qwen3-VL text encoder.

2026-09-01 ~ 2026-09-02 · 2 related posts

Full story(15 episodes)→