Voice Bank
Cloned voices, their samples and their delivery settings
Voices
A voice is one audio sample plus a set of delivery settings. The sample is re-uploaded with every synthesis request, so its size matters across a long project.
Loading voices…
No voices yet
Upload a clean MP3 or WAV of the speaker you want to clone — 30 to 60 seconds of steady narration is plenty. Everything else on this page is tuning on top of that one file.
Add voice
Voice sample
required
Drop an MP3 or WAV here, or click to browse
Up to 7 MB. Under 2 MB keeps a long project cheap.
Free text. It becomes “Narrate in a <category> style.” in the prompt.
Passed to the model almost verbatim, so write it as an instruction.
0.97
45
75
9
MIMO exposes no numeric knobs. These four sliders are translated into a style prompt, and that prompt is the only thing the model actually receives.
Preview
Trimmed to 800 characters. Long enough to hear pacing, short enough to be quick.
This calls the cloning model and uploads the full sample, so it takes several seconds.
The button stays busy until the audio comes back.