Choose a voice, write your script, and shape the delivery. A blank line between paragraphs adds a longer pause.
Create a voice from recordings, manage reference clips, and build or refine voice profiles. Record only voices you own or have permission to use.
Record or upload 1–5 clean WAV samples (10–30 seconds, only this speaker) so Studio can identify their voice in videos. Then upload videos to extract and filter that speaker’s clips.
Read at a comfortable pace in a quiet room. Each take is converted to WAV in your browser, previewed, then saved on this computer as a voice sample. Four original prompts help capture varied sounds and delivery.
Review one clip at a time. Delete removes the source clip, Keep saves your decision without promoting it, and Promote adds it to this person’s active six in place of the weakest reference. Prepared datasets use their speaker-match scores; folders use audio quality.
The highest-ranked unreviewed clips from the selected sources. Preview any clip or promote it directly.
Build a candidate, then generate both profiles from the same short script, settings, and seed. Listen before adopting it.
Play the takes, keep the best one, and mark approved when it is ready to use.