An AI song cover is the same song with a different singer. Nothing gets rewritten: the notes, the timing and the arrangement stay, and only the voice changes.
Upload a song as audio or video, up to 6 minutes. Demucs splits the vocals from the instrumental, then the YingMusic singing voice conversion model sings the vocal part again in the voice you picked. The new vocal is mixed back over the original backing track, and the loudness of the final track is evened out.
There are six singers, all AI-generated: a bright pop soprano, a warm breathy alto, a powerful raspy rock belter, a soft pop tenor, a deep velvety baritone and a gritty male rock voice. Listen to each one before you start. To hear the song in your own voice, upload 8 to 60 seconds of yourself without music; singing works best. You can push the vocal up or down in the mix and add a light or hall reverb.
If you like, AI also paints square cover art, which is embedded in the MP3. A 2–3 minute song takes a few minutes.
Use songs you wrote, tracks you have the rights to, or make a cover just for personal listening; you confirm this before starting. There are no famous singers' voices here and there won't be: you can only clone your own voice or the voice of someone who has agreed to it.