FastH3-Omni on Reactor - video with synchronized audio, from a single prompt.
Your daydreams have a sound department now.
MiniMax's FastH3-Omni generates synchronized audio video from one prompt
AIMiniMax's FastH3-Omni, an open-source realtime H3 model post-trained by FastVideo and Nuva Lab, generates video with synchronized audio from a single prompt on Reactor. The model supports first and last frame conditioning, audio input with dialogue, and up to three reference images.
@MiniMax_AI
Introducing FastH3-Omni. The fastest and highest-quality open-source version of realtime H3 by @MiniMax_AI, post-trained by FastVideo @haoailab with @nuvalab. Exclusive preview on the Reactor API until 10.20.2026. • First and last frame conditioning • Audio input + strong dialogue • Up to 3 reference images • Frontier-level performance
Source details
MiniMax (official) · x.comPublished · added here
Original title: FastH3-Omni on Reactor - video with synchronized audio, from a single prompt.