MiniMax open-sourced H3 in early August — 15-second 2K clips with native stereo audio. Three weeks later comes the real move: MiniMax Design, a desktop app (macOS 13+, Windows 10+, free to start) that wraps H3 in a full production pipeline.
Four agents, one canvas
Describe the video you want and four agents start in parallel: a copy agent writes the script, an image agent generates stills, a video agent assembles the timeline, an audio agent handles voiceover and music. Brief, storyboards, H3 clips, and edits all live on one canvas. There’s even a 3D director booth for placing characters and setting camera angles before you burn credits on generation. A Skills plaza lets you install workflows built by creators — short drama, e-commerce, commercial ads — or build your own through chat.
The API side
Design itself is a desktop app, but H3 sits in MiniMax’s open platform API alongside its text and speech models. Developers can script the same video generation into their own pipelines.
The bigger story: another Chinese model lab deciding that selling tokens isn’t enough — the agent workbench is the product now.
You Might Also Like
- Figma Code to Canvas Just Flipped the Design dev Workflow on its Head
- Google Releases Gemini Embedding 2 one Vector Space for Text Images Video and Audio
- Design Agent by Lokuma Topped Product Hunt With a Simple Pitch ai Agents Need Designers too
- Sora is Dead ltx 2 3 Lightricks Ships 22b Open Source Video Audio in a Single Forward Pass
- Insanely Fast Whisper Hits 3 on Github Trending 150 Minutes of Audio Transcribed in 98 Seconds Zero Cloud
Top AI Product 
Leave a comment