MiniMax H3 Max Ref-to-Video: turn your references into video

MiniMax H3 Max Ref-to-Video is the reference-driven member of the MiniMax H3 Max family. Instead of starting from a blank prompt, you feed it multiple reference images, video clips, and audio files, then describe how to use them. It generates a new clip of 5 to 15 seconds at 480p or 768p, in adaptive or fixed aspect ratios from 21:9 to 9:16 - grounding the result in the material you provide.

Ref-to-Video reads the references you attach in order and lets you point to them directly in the prompt - Image 1, Video 1, Audio 1, and so on. Pull a character from one image, a motion or setting from a video clip, and a voice or soundtrack from an audio file, then describe how they should come together. The model composes a single coherent video that honors each reference, so your source material shapes the shot instead of a text description alone.

Ref-to-Video is built for control: attach as many images, clips, and audio files as your shot needs, then use the prompt to direct exactly how each is used. Choose 480p or 768p, set a length from 5 to 15 seconds, and pick an adaptive ratio or a fixed one from 21:9 to 9:16. The result stays grounded in your material while the model handles motion and continuity - so your references become a finished video, not a mood board.

MiniMax H3 Max Ref-to-Video is available in Picsart's AI Playground, where you upload references, write your prompt, and generate - then compare the result against 150+ other AI models from a single prompt, with no setup required. Want the fastest text- and image-to-video instead? MiniMax H3 Max Turbo; need native 2K with audio? MiniMax H3. And you can reach it whichever way you work: on the web, in the Picsart desktop app, or built into your own projects via CLI, MCP, REST API, and SDK.