The omni-modal video model that reads text, images, video and audio as one context. Generate a shot, swap a character, change a line — without touching the rest.
Why MiniMax H3
Reference up to 12 files in a single prompt — 9 images, 3 video clips, 3 audio tracks. Lock a face, borrow a motion, set the pace to a soundtrack. H3 reads them as one brief, not a collage.
Reference a face, an action, a soundtrack — in one prompt.
Every output ships with native stereo audio — dialogue, ambience and effects generated with the picture, not dubbed over it afterward.
No silent drafts. No syncing in post.
Swap a character, change the background, rewrite a line of dialogue. H3 executes local edits with precision — and keeps everything you didn't touch stable.
Only the instruction changes. Everything else stays.
Brand films, product showcases, story scenes, stylized animation — cinema-quality frames where captions, logos and product details stay exactly where you put them.
From brand film to product demo — mix H3 with Seedance or Kling in one timeline.
See It in Action
Brand Film
Brand campaign with locked product identity and a native soundtrack.
Product & E-commerce
Same SKU, three scenes, zero drift — shape, material and logo hold.
AI Drama
Multi-cut dialogue scene, generated with sound — lip-sync included.
Stylized Animation
A real street scene, restyled — motion and light preserved.
Specs
How It Works
Tell Pixo's agent your idea. Pick aspect ratio and resolution right at the prompt — then it drafts your script and full storyboard.
Tell the agent you want MiniMax H3, or let it choose the right model. Your references ride along, and every shot is generated in sequence.
Reorder shots on the timeline, fine-tune the ones that need it, and export the finished video — watermark-free.
FAQ
MiniMax H3 is a new-generation omni-modal video model from MiniMax, the team behind Hailuo. It treats text, images, video and audio as a single context, and generates 4–15 second multi-shot videos with native stereo audio. On Pixo, it sits alongside Seedance, Kling and Veo in the same storyboard.
Videos of 4–15 seconds at 24 fps, in six aspect ratios from 21:9 cinema wide to 9:16 vertical — chosen when you write the prompt. Every result comes with sound built in.
Text prompts up to 7,000 characters, plus reference files — images to lock a face or a scene, video clips for motion and camera moves, and audio tracks for pacing and voice.
Yes. In Pixo's storyboard, every shot can use a different model — mix MiniMax H3 with Seedance 2.0, Kling 3.0 or Veo 3.1 in the same timeline.
Sign up at pixo.video — new users get 200 credits on sign-up, and plans are currently up to 55% off.
Yes. Generate in 768p to iterate quickly, then upgrade the result to 1440p — the officially recommended mode for final output. A 21:9 video renders at 2976×1248 in 1440p mode.
Up to 12 files: 9 images, 3 video clips totaling 15 seconds, and 3 audio tracks. Audio must be paired with an image or video input. H3 reads them all as one context — lock a face with an image, borrow a motion from a clip, set the pace with a track.
Generated speech covers 11 languages with precise lip-sync — including English, Chinese, Japanese, Korean, French, German and Spanish — and 40+ further languages such as Arabic, Thai and Hindi can be explored. That makes it a fit for multilingual ads and localized product videos.
New users get 200 credits on sign-up. Plans up to 55% off.
Try MiniMax H3All models. One storyboard. Export watermark-free.