Loading your voices…
Train your voice — no song needed
Build a private voice model from a short reading script — about six to nine minutes, no singing, any microphone. Then re-voice any take in the Studio: the performance stays yours, only the timbre changes.
Already have a recording? Record or upload a 30–60 second sample below.
Training your own voice is part of Pro
Voices is where you build your Personal Trained Voice — a real human voice model that makes the whole song human. On the Starter plan your AI vocals are still fully humanized and the instrumental is rebuilt with real instruments; upgrade to Pro to add your real voice, train as many voices as you like, and blend a real human voice into the song.
Upgrade to Pro →A voice is learned from a short sample of your voice and stays private to your account.
Line input — browser processing (AGC / noise suppression) is off.
Read a short script and hum a few vowels — we stitch it into one clean training sample.
One private voice, plus unlimited hybrids from the Collaboration Studio. The Voice Map matches sections automatically.
Record or upload a clean 30–60 seconds of you singing or talking — no music, no other voices.