Upload a short reference recording, enter what it says, and synthesize new speech in the same voice.
A clear 3–10 second clip works best. Maximum upload: 25 MB.
Leave blank to transcribe automatically; the first request will take longer.