# Text to speech

## Choose a source voice

Use **TTS** to preview the script with a voice in its language. Current releases offer online Edge voices and installed Windows voices for offline use. Preview names and punctuation, and adjust pacing before passing the result through RVC.

## Convert the generated speech

Select the target inference model and matching embedder. Start at zero transposition and check the output against the unconverted TTS generation. A pronunciation error present in both files belongs to the TTS stage. A new pitch jump or distorted consonant belongs to the conversion stage.