Twilio

Twilio Media Streams pipes real-time call audio to Deepgram for transcription. The reference architecture forks audio from calls to a proxy server via WebSockets, the server forwards to Deepgram, and transcripts flow back to subscribed clients. Setup needs a Twilio number, a Deepgram API key, and ngrok for local development.

• Transcribing Twilio Voice Calls in Real-Time with Deepgram
• Transcribe Twilio Voice Calls in Real-Time with Rust and Deepgram
• Twilio and Deepgram STT (developer docs)
• On-premise Twilio integration tools

For batch transcription of completed calls, Programmable Voice recordings flow into Deepgram via API.

• Transcribe Recorded Calls With Twilio (developer docs)

Aura-2 handles speech synthesis for AI-generated responses on the outbound side of a Twilio call.

• Twilio and Deepgram TTS (developer docs)

Full voice agent flows orchestrate STT, an LLM, and TTS through Twilio. Documented patterns include OpenAI as the LLM and the Deepgram Voice Agent API as the speech layer.

• Build a Voice Agent with Twilio, OpenAI, and Deepgram (developer docs)
• Twilio and Deepgram Voice Agent (developer docs)

For non-developer ops teams, Twilio Studio plus Pipedream plus Deepgram delivers automated transcription pipelines with email or webhook delivery.

• Email Phone Call Transcripts with Twilio Studio and Pipedream

If you are building voice features on Twilio, the fastest start is the Marketplace Add-On. For the integration patterns above, the developer docs walk through each end to end. Enterprise terms via deepgram.com/contact-us.

• Twilio Marketplace Add-On for Deepgram (find Deepgram listing)
• Twilio Media Streams documentation
• Deepgram documentation
• Contact Deepgram