Deepgram, the speech recognition company, has integrated its technology directly into Together AI's platform, allowing developers to build voice-powered applications without juggling multiple providers.
The partnership means Deepgram's speech-to-text engine now runs natively inside Together AI's infrastructure, keeping audio data, AI model processing and activity logs in one place rather than routing them across separate services.
That matters most for real-time voice agents, the kind used in customer service and healthcare, where even small delays break the flow of conversation.
Developers using Together AI can now select Deepgram as their transcription layer within existing pipelines, rather than connecting an external service and writing additional code to make the two systems talk to each other.
Everything is billed through a single account, and full transcripts are retained for quality checking and call routing.
Related reading
- Deepgram highlights women building the future of artificial intelligence
- Deepgram identifies five pronunciation error types costing voice AI deployments millions annually
- Facebook Marketplace adds Meta AI selling tools
Deepgram says the co-located setup reduces the time audio spends travelling between systems, which in turn speeds up how quickly a voice agent can respond to a speaker.
Teams that need tighter data controls can also combine the Together AI integration with Deepgram's own dedicated or self-hosted environments, which the company says supports compliance and multi-region requirements.
The recap
- Deepgram STT is now hosted natively on Together AI pipelines.
- Live demo available by calling (847) 851-4323, the company said.
- Developers can plug Deepgram into Together AI via voice quickstart.