AI music video production

The song comes first

A song cannot simply be laid over finished footage. Where the shot changes, how a scene unfolds between verse and chorus, and which character sings a given line are all decided before generation. That is why the storyboard ties planned frames to parts of the track, and why arguing about the drama is cheaper at that stage.

Beat-aware editing

Shot boundaries are tied to musical beats, and a cut is placed in the pause between vocal phrases. When one beat grid runs through the whole project, the edit stops being a rescue operation. Videos assembled from separately generated good shots fall apart exactly here: the song has rhythm and the frames have rhythm, but there is no shared grid.

Who sings on screen

For videos with singing characters the vocal is aligned to the lyrics, and lip sync is driven by instructions with timing. Lead vocal, backing vocals and ad-libs are told apart, so lines can be assigned to specific characters: someone sings on screen while someone else stays out of frame. Lip sync can also be switched off, which makes it a choice between two genres.

Two genres of music video

A performance video rests on expression and articulation: the singer is on screen, close ups, light. A story video runs a plot with the song in the background, where the scene and the cutting rhythm matter more than lip movement. Mixing the two modes inside one minute usually reads as chaos, so the mode is picked before the storyboard.

How the work runs

It starts with the brief and the track. Then the song breakdown, the storyboard and a character reference sheet, so the face and wardrobe stay put between shots.

• brief: track, idea, where it will run
• song breakdown: sections, timings, vocalist
• storyboard with frames tied to parts of the track
• character reference sheet and a style test
• frame generation with a review of every frame
• beat-aware edit, sound, export and a sync check in the file

Rights to the track and the vocal

Rights to the track and the vocal stay with you: we work with the material you provide. Music for our own videos we compose ourselves, so no stock tracks end up in the frame. We do not generate recognisable characters or franchise attributes: those are someone else's property, and permission comes from the rights holder.

The music video and neighbouring services

When every frame has to be worked by hand, with precise acting, that is handcrafted video. Vertical cuts of the video for feeds and testing are covered by vertical ads, and versions across accounts and geos by mass posting. A regular run of episodes for a channel is set up as an AI YouTube channel. A story-driven cartoon with a character is a separate cartoon production page.