Singing Voice Changer
Re-sing a vocal in the voice of any sample clip. Melody, timing, and words stay; only the singer changes.
Singing Voice Changer: Any Vocal in Any Voice
The Singing Voice Changer takes a vocal recording and a short clip of another voice, and returns the same performance sung in that voice. Melody, timing, and words stay exactly as recorded; only the singer changes. No voice library, no ids: the sample you upload is the target.
It runs on Seed-VC, an open zero-shot singing conversion model with pitch conditioning, hosted by mitte.
What can you do with it?
- Replace the singer on a demo from one clean sample of the voice the release needs
- Hear a topline in another voice before booking a session
- Turn one take into a duet or a choir by converting copies into different voices
- Sing in a voice you cloned by using its sample as the target
- Re-voice a reference vocal so the guide track is not the writer’s voice
How to use it
- Upload your vocal. An a cappella or a vocal stem (mp3, wav, flac), up to 10 minutes. For a full song, run the Vocal Splitter first and use its vocals stem.
- Upload a voice sample. Up to 30 seconds of the target voice, sung or spoken, one voice, no music behind it.
- Run it. A song takes a couple of minutes; the first run after a quiet spell adds about a minute of warm-up.
- Get your vocal. One labeled audio output in the new voice, ready to lay back over the stems.
Tips
- Feed it vocals only. Instruments in the vocal confuse the model; that is what the Vocal Splitter is for.
- A sung sample matches better than a spoken one, and a clean one better than a noisy one.
- Shift the pitch when the target sits in another range. A baritone song into a soprano voice usually wants a few semitones up; find it in the settings.
- Only use voices you have the right to use. Your own, a collaborator’s with their consent, or a licensed one.
- Effects go on after. Convert the dry vocal, then add reverb and compression in the editor.