tts

# Text to Speech ## When to use Reading text aloud or rendering speech audio files, using only the voices already on the machine. ## Setup No account or API key. Pick the platform command; fail loud when the Linux engine is missing (`sudo apt install speech-dispatcher` or `espeak-ng`): - macOS: `say -v Samantha -o out.aiff "text"` (list voices: `say -v '?'`) - Linux: `spd-say "text"` or `espeak-ng -w out.wav "text"` - Windows: PowerShell `System.Speech.Synthesis.SpeechSynthesizer` ## Workflow 1. Keep utterances short; split long text at paragraph breaks. 2. Confirm voice and speed when the user cares; otherwise default voice, normal rate. 3. For files, report the path, format, and duration. ## Non-goals - Do not clone voices or impersonate a real person. - Do not call cloud TTS APIs (different cost and privacy posture — ask first). - Do not play audio unprompted in shared environments; render a file instead.