Deepgram Acquires Pipecat-Maker Daily to Build Open Voice-AI Infrastructure
Written by AI from techcrunch.com's report. Not checked by a person.
Enterprise voice AI company Deepgram acquired Daily, the startup behind the popular open-source Pipecat framework that has raised over $62 million in funding, pledging to keep the project open-source and launching a new open protocol for real-time voice agents.
Deepgram announced on Thursday that it is acquiring Daily, the startup that created and maintains Pipecat — one of the most widely used open-source frameworks in voice AI. The deal's financial terms were not disclosed, but both companies have committed to keeping Pipecat open source.
Pipecat is a framework that manages real-time voice and video streams alongside components such as speech-to-text, large language models, and text-to-speech, all while reducing latency. Daily also provides hosted infrastructure for voice and video calls. The company has raised over $62 million from investors including Tiger Global, Highfield Capital, Nat Friedman, Lachy Groom, and Kevin Weil. Deepgram, for its part, reached unicorn status earlier this year with a $130 million Series C round. Notably, both companies graduated from the same Y Combinator batch in 2016.
Daily co-founder Kwindla Kramer said the Pipecat framework emerged from experiments in AI-to-human communication over voice and text that began around the time GPT-4 launched. He said he was drawn to Deepgram CEO Scott Stephenson's personal commitment to keeping Pipecat open source and vendor-neutral permanently, and that the engineering cultures of both organizations were a strong match.
Stephenson said Deepgram had long wanted to engage with open source and identified an orchestration framework for voice agents as the right entry point. He stressed that he wants developers to use whichever models best fit their needs, even competitors'.
Alongside the acquisition, Deepgram is launching the Real-Time Context Standard (RCS), an open protocol for voice agents co-developed with the Pipecat team. RCS defines metadata including audio context signals — such as detecting a whisper — instructions for a language model, and guidance on how a response should sound.
