MiniMax 将在 ModCon 介绍 H3 AI 视频生成能力

ModCon 2026 is August 18th at the Grand Hyatt in San Francisco, and is our flagship developer conference for the people building, running, and

H3 can replace, add, or remove characters and objects; change backgrounds, lighting, atmosphere, and visual effects; and revise dialogue, human voice, or voice identity. Its precise instruction-following is designed to apply targeted changes while keeping unedited content relatively stable. H3 generates videos from 5 to 15 seconds at 24 FPS, and every result includes native stereo sound. It supports common ratios including 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16, with 1440p output available. The product documentation lists 768p as a later rollout option. [...] H3 supports Text-to-Video, first-frame and first-and-last-frame generation, and an Omni Reference workflow. Omni Reference accepts up to 9 images, 3 video clips, and 3 audio clips, with a combined limit of 12 files. Audio references must be used together with an image or video. You can reference character appearance, actions, camera movement, composition, voice, visual atmosphere, narrative rhythm, sound design, and editing style.

Morgan Suo, Head of US Business Development, joins ModCon '26 to share what open-weight omni-context generation enables for AI video.

Join us at ModCon on August 18, 2026 in San Francisco for a full day of AI infrastructure launches, technical deep dives, hands-on demos,

Watch the full ModCon 2026 keynote, recorded live at the Grand Hyatt in San Francisco on August 18, 2026.

4 months ago, we could see things are changing every month, every few weeks. Now for every single episode, things are changing dramatically. Dramatically, yeah. Like the difference is not uh you know small percentage. from a minute to 50 seconds, a minute to 6 seconds. And I think for people who are watching the show, this is great that you guys are keeping yourself up to date even if you are not trying. Even if you are not trying, if you are not going to the file, knowing that what is happening and then applying it to different things. This is happening to the video today. It happens to image. It happens to code. It happens to whatever industry you are in. So, things are changing really fast and it's really important to understand how to adapt, how to change your business process. It [...] But the quality of the videos has actually improved in terms of prompt adherence compared to the original model. That takes much longer to generate the videos. Look at that third row there. Average