Local CPU TTS (Piper). No login.
If set, this overrides the dropdown voice for Elevate.
XTTS uses your uploaded sample to clone the voice.
Tip: first run downloads a few voices (~hundreds of MB total). Generation is CPU-only.