AnalysisAI ModelsAugust 5, 2026

Local MiniMax H3 renders 2-minute uncensored video

A Reddit test ran MiniMax H3 locally, using 24.5 GB of VRAM and ~8.5 minutes per shot on a Runpod GPU, costing about $3 for the full two-minute video. Rendered via ComfyUI, it avoided cloud censorship.

1 source

Daily brief

Get tomorrow's AI brief in your inbox

More stories today

Open the live feed