Local Speech-to-Speech

Mic → faster-whisper → Ollama → XTTS voice clone · 100% offline after setup

Cloned voice (input_voices/)

Ollama model

Language

Whisper size

Drop reference clips into input_voices/ then refresh. Soft/close-mic female clips clone best for intimate tone.

Optional typed message (skips mic if filled when you click Speak Reply)

Ready.