MiniMax H3 image-to-video starts from your opening frame, with an optional closing frame and a prompt that directs the motion between them. Clips run 4 to 15 seconds with generated audio at 768p or 2k.
Workbench
These examples were written for this endpoint. Replace the prompt and sample media URLs with your own accessible inputs, then check the current quote and model availability before submitting.
- Output7s
cup-then-two-in-one-shirt The opening frame is a still life; everything the prompt asks for happens after it. A hand lifts the cup out of frame, the camera tilts and racks focus, and a second figure appears deeper in the room — three changes, all of them temporal, none of them statable in the input image. That is the split this endpoint rewards: `image_url` owns the composition, and the sentence owns the change.
- 768p
- 6s
- Output5s
windowsill-turn One opening frame plus a movement note. Everything identifying - wardrobe, face, room - is decided by the frame, not the words.
- 768p
- 5s
- Output8s
robe-to-doorway Both ends pinned, indoors: `image_url` is the opening moment, `last_image_url` is the doorway turn-back she reaches over `duration_seconds` `8`. The `prompt` covers only the walk between — a slow handheld push and footsteps on wood — since the stills already answer what she wears and where she lands. Generate a start and end pose, let the model fill the rest.
- 768p
- 8s
This example starts at $0.037 per output second. The Playground updates the estimate when you change the settings.