
MiniMax H3
A general multimodal video model: text, image, video, and audio in, 15-second 2K clips with stereo out.
Visit MiniMax H3 Pay per generation via MiniMax and partner platforms
Quick verdict
- Best for
- Restyling or re-animating existing footage
- Pricing
- Pay per generation via MiniMax and partner platforms
- Not ideal if
- 15 seconds is half what Seedance 2.5 and Wan3.0 offer
What is MiniMax H3?
MiniMax launched H3 on July 31, 2026 as a general multimodal video model that takes text, image, video, and audio context and generates up to 15 seconds at 2K with native stereo sound in the same pass. It also lists video-to-video motion transfer, multimodal editing, and multi-shot generation, and is distributed through MiniMax's own apps and partner platforms including Luma Agents.
Key features
- Text, image, video, and audio input
- Up to 15 seconds at 2K with native stereo
- Video-to-video motion transfer
- Multimodal editing and multi-shot generation
- Available through partners such as Luma Agents
Pros
- Motion transfer and editing make it more than a prompt-to-clip tool
- Native audio avoids a separate sound pass
- Broad partner distribution
Cons
- 15 seconds is half what Seedance 2.5 and Wan3.0 offer
- Access terms through partners vary
- Chinese-hosted model considerations
Best for
Restyling or re-animating existing footageMulti-shot social clips with soundUsers of platforms that bundle H3
Read more
Related comparisons
Alternatives to MiniMax H3
Hailuo AI
MiniMax's AI video — free, fast, surprisingly good.
FreeFree
Released September 2024Seedance 2.5
ByteDance's 30-second audio-and-video model with timeline editing of individual sections.
PaidVia Jimeng AI and Doubao Pro subscriptions; per-generation pricing on partner platforms
Released July 2026LTX-2.5
Open-weight video model with synchronised audio, native multishot, and 4K HDR output.
FreemiumOpen weights free to run; hosted generation priced per second in LTX Studio and partners
Released August 2026

