voice

---

name: voice description: Voice integration — TTS via OpenAI tts-1, STT via Whisper, audio caching. Trigger keywords: voice, tts, stt, text to speech, speech to text, whisper, audio. metadata: author: CorvidLabs version: "1.0"

Voice — TTS & STT Integration

Text-to-speech and speech-to-text capabilities using OpenAI APIs.

Architecture

• TTS: server/voice/tts.ts — OpenAI tts-1 model
• STT: server/voice/stt.ts — OpenAI Whisper API
• Gated behind: OPENAI_API_KEY env var
• Cache: voice_cache database table (hashes text, avoids re-synthesis)

Text-to-Speech (TTS)

import { synthesize, synthesizeWithCache } from '../voice/tts';

// Direct synthesis (no caching)
const audioBuffer = await synthesize(text, voice);

// Cached synthesis (checks voice_cache table first)
const audioBuffer = await synthesizeWithCache(text, voice);

Available Voices

OpenAI tts-1 voices: alloy, echo, fable, onyx, nova, shimmer

Caching

synthesizeWithCache() hashes the input text and voice, checks the voice_cache table, and only calls the API if the audio isn't cached. This saves API costs for repeated phrases.

Speech-to-Text (STT)

import { transcribe } from '../voice/stt';

const text = await transcribe(audioBuffer, 'audio/webm');

Calls the OpenAI Whisper API to convert audio to text.

Telegram Voice Notes

The Telegram bridge automatically:

1. Receives voice messages from users
2. Transcribes them via STT
3. Processes the transcribed text as a normal message
4. Can optionally respond with a voice message via TTS