Your characters, on stage

An AI music video generator turns a piece of music into footage of someone performing it. Instead of animating a single face, it stages a performance: the singer's lips follow the vocal, the bodies move with the rhythm, and the camera cuts and travels the way a music video's would.

Mage's Music Video app is built around characters. Choose one to three performers from your saved characters, add audio of up to 15 seconds, and optionally add up to three reference images for the stage, the wardrobe or the props. Mage films the performers singing and dancing in sync with cinematic camera work, on Plum by default, or on Plum Max, Lemon or one of the four Cherry models.

Because the performers are saved characters, the same faces appear in every clip, which is what makes a run of short clips feel like one video. Make the chorus, then a verse, then the bridge, and put them in order in Storyboard.

Each clip takes up to 15 seconds of audio, so plan the song the way an editor would: intro, verse, chorus, bridge, final chorus. Cut the audio at the start of each section, generate a clip per section with the same performers, and vary the prompt so each section has its own setting or shot. A chorus on a rooftop and a verse in a diner reads as a real edit.

When the sections are done, Storyboard lines up to six saved clips in order and exports one MP4 with their sound, in 9:16 or 16:9. Six clips of 15 seconds cover a minute and a half, enough for a short-form cut of most songs.

• Cut the audio at section boundaries, never mid-word
• Keep the same performers in every section
• Change the location or the lighting per section
• Use reference images so the wardrobe stays consistent
• Choose 9:16 for short-form feeds, 16:9 for a widescreen cut

Performers are saved characters, so their look is set before you open the app. Build each one from a clear, front-facing portrait; if you are designing them from scratch, the AI character generator walks through making one and saving it. The Character Sheet app turns one image into front, side and back views, which is a quick way to check a design before it goes on stage.

For a band, keep the styling in one family: the same palette, the same era, the same wardrobe rules. Add the outfit or the stage as a reference image rather than describing it at length, and the clips will agree with each other.

When the vocal matters most, a close shot of one performer gives the cleanest sync. For dancing, a wider shot with room around the bodies works better, and you can name the style of movement in the prompt. A duo trading lines works best when the prompt says who sings when.

For choreography copied from a real dance clip, use Motion Control instead, which takes the movement from a reference video. The AI dance video generator guide explains both routes, and the AI music generator shows how to make a track from a prompt if you do not have one yet.