MiniMax (official)· @MiniMax_AI · X·· 2 小时前AI 评分52
AI 导读
MiniMax 宣布 FastH3-Omni 登陆 Reactor，可从单个提示词生成带同步音频的视频，10.20.2026 前在 Reactor API 开放预览。FastVideo 与 nuvalab 对 MiniMax 的实时 H3 做后训练得到该开源版本，支持首尾帧条件控制、音频输入与强对话能力、最多 3 张参考图像。
正文
FastH3-Omni on Reactor - video with synchronized audio, from a single prompt.
Your daydreams have a sound department now.
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在 X 查看被引用的帖子
来源：MiniMax (official) · x.com