The input is a URL to a track you already have — an upload or a prior generation — plus a prompt that describes the new arrangement. The cover job reinterprets that song; it does not write a new one from a blank brief. An instrumental flag drops the vocal if you only wanted the restyle.
Text-to-music is the original generate, and that page already warns against artist-clone prompts. Stem-separation splits a mix you own; it does not restyle it. Rights in the source still sit on usage-rights: generating a cover does not confer a licence to a third-party original.
Lay the result under picture the same way you would any other bed. The editing task and the agent job are the same object in two doors.
In practice
- Point at a track you have rights to before you describe the new style.
- Ask for acoustic, lo-fi, synthwave — arrangement language — not 'in the style of a named artist.'
- Use instrumental when the cover has to sit under a voiceover rather than compete with one.
The mistake to avoid
Prompting a blank generate for a cover of a song you never pointed at, or assuming the cover cleared the source rights.
Go deeper
Cover Versions of Tracks You Already Own
An AI cover doesn't compose a new song — it reinterprets a track you supply. That single mechanical fact is why the whole question is source rights.
Where you will run into it
- Add a Cover Song to Your Video — Same song, reimagined arrangement.
- AI Music Generator — Describe a vibe. Get a song. Keep the rights.
Related terms
Text-to-music
Text-to-music generates an original composition from a written description of genre, instrumentation, mood and tempo — with or without sung lyrics.
Stem separation
Stem separation splits a finished mix into its component parts — vocals, drums, bass, other instruments — as separate audio files.
Usage rights
Usage rights are the contract terms that say where a brand can run a creator's content, for how long, and in what form — organic-only, paid, whitelisted, in perpetuity — and they default to the narrowest reading if the contract doesn't say.
Audio extend
Audio extend continues a previously generated music track from a chosen point, rather than writing a new song or time-stretching the file.
Text-to-speech
Text-to-speech converts written text into spoken audio using a synthetic voice you choose before generating.
The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync in a single app.