Loading supported attributes…
A few seconds of clean speech works best.
Generation parameters
Set seed to any number to get the same audio every time. Position
temperature 0 makes sampling deterministic; note it does not fix the
non-verbal tag failure — nothing does, so spread tags out instead.
The same request, in code
Updates as you change the form.
Available voices
Loading…