# FastH3 Omni: an open, much faster version of MiniMax's H3 video-with-audio model.

> On Oct 20, FastVideo (UC San Diego's Hao AI Lab) and Nuva Lab released FastH3 Omni.

Published 2026-10-20T17:28:39-07:00 · News · Canonical: https://news.littlewolflabs.com/posts/2026-10-20-fasth3-omni

On Oct 20, FastVideo (UC San Diego's Hao AI Lab) and Nuva Lab released FastH3 Omni. It's a slimmed-down, faster version of MiniMax H3 that makes 5 to 15 second videos with synchronized sound from a prompt plus up to nine images, three video clips and three audio clips. It needs 8 generation steps instead of 50. On eight GB200 server GPUs, the team measured it finishing 5.6 to 11.9 times faster than the original. The weights are on Hugging Face, with ComfyUI versions, home setups for an RTX 5090 or DGX Spark (480p by default) and a lower-precision version that runs on an RTX 4090. It's free to download, but the weights fall under the MiniMax-H3 Community License, so read it before any commercial use. *Try it:* it's worth testing if you have a 4090 or 5090. The speed numbers are the team's own and were measured on server GPUs, not home cards. Sources: [FastVideo](https://haoailab.com/blogs/fastvideo-omniref-update/), [Hugging Face](https://huggingface.co/FastVideo/FastVideo-FastH3-Omni-8-Step-V1), [Nuva Lab](https://nuvalab.ai/fasth3/).