In plain words
MiniMax H3 is a video model from MiniMax. It makes clips from a description, from a photo, or from a first and a last picture, filling in the movement between them.
Clips last from 5 to 15 seconds, come with sound, and are made at 2K, a bit sharper than Full HD.
How to get the most out of it
- Say what moves and in what order. “Buds open, then petals drift down” reads better than “a tree in spring”.
- Name the sounds you want. “Birdsong”, “waves”, “an accordion”.
- For two pictures, keep the camera in the same place. Only the thing that changes should be different.
- Start from a clear, sharp photo. A blurry start gives a blurry clip.
How it compares with the rest of the family
MiniMax H3 is MiniMax’s video model. Its chat models are Minimax M3, which thinks before it answers, and the M2 line: Minimax M2.7, Minimax M2.5 and Minimax M2.1. A chat model can help you write the scene description before you make the clip.
When to pick another model
For 4K, try Veo 3.1 or Kling V3.0 I2V. For longer clips with sound, Flux 3 goes up to 20 seconds and Seedance 2.5 up to 30.
