One song becomes one short film.

The song's length is the video's length

Music generation has no length parameter — the lyrics decide how long the song runs. So the song is confirmed first and the video is cut to fit it, not the other way around.

Scenes and clips are different things

A scene is a unit of story. A clip is the 4 to 15 seconds the video model can make in one pass. A long scene is split across several clips, and those clips share the scene's beat and references.

A shape for the story

You pick the skeleton before anything is generated. The shape you choose follows every scene and sets how hard each one pushes.

• Arc
• Return
• Reveal
• Intercut

People and places stay put

Character sheets and background sheets are built first and fed into every clip as references, so the same person does not arrive with a different face and the same place does not become somewhere else.

Clip stills are skippable

A place to see the framing as a picture before spending on video. The sheets are composited in, so faces and places match them, and the still you made rides along as a reference when the clip is generated. Skip it if you do not need it.

Landscape 16:9 only

There is no vertical music video. Video is fixed to MiniMax H3 at 768P, so there is no model to choose.