Deepgram Voice AI is buying Daily, the company that builds and maintains the widely used open-source Pipecat framework, in a move meant to reinforce its real-time voice and video stack. The firms announced the deal on Thursday without naming a price and said Pipecat will stay open source.
Daily hosts infrastructure for voice and video calls. Pipecat sits in the middle of a voice agent pipeline: it orchestrates live audio and video streams together with speech-to-text (turning speech into text), large language models, and text-to-speech (turning replies back into speech), while aiming to cut delay. With that toolkit inside its fold, Deepgram Voice AI expects a sharper grasp of how real-time communication systems are built end to end.
Funding, shared roots, and why Daily said yes
Daily has raised more than $62 million from backers including Tiger Global, Highfield Capital, Nat Friedman, Lachy Groom and Kevin Weil. Deepgram reached unicorn status earlier this year after a $130 million Series C round. The two companies also shared a starting point: both were in the same Y Combinator batch in 2016.
Kwindla Kramer, Daily’s co-founder, said the firm began testing AI-to-human talk over voice and text around the time GPT-4 appeared, work that led to Pipecat. He said other buyers had shown interest, but he preferred Deepgram chief executive Scott Stephenson’s promise to keep Pipecat open source for good, and felt the engineering cultures lined up. That match, Kramer said, should let Daily’s team work across more of the stack inside a larger company. Stephenson, he added, was “super committed to Pipecat being open-source and vendor-neutral forever,” seeing openness as both a personal stance and a commercial edge in a fast-growing ecosystem.
What Deepgram Voice AI wants from the deal
Stephenson said Deepgram had long wanted a stronger open-source presence and saw agent orchestration as a gap that made a Daily partnership attractive. The aim, he said, is to open the ecosystem so developers can build voice agents with whichever models fit their needs—not only Deepgram’s—while Deepgram improves where partners already lead.
The acquisition, he said, should help builders assemble expressive, end-to-end speech systems by swapping models and using a “context bus” that links the pieces. Because Pipecat also handles real-time media beyond pure voice, Deepgram Voice AI can probe related formats of live communication, not only speech alone. Kramer framed Pipecat less as a narrow voice kit and more as a toolkit for AI-native software that assumes real-time interaction from the start.
New open protocol and a busy market
Alongside the purchase, Deepgram is launching Real-Time Context Standard (RCS), an open protocol for voice agents co-developed with the Pipecat team. RCS describes metadata such as cues drawn from audio (for example a whisper), instructions telling a language model how to use those signals, and guidance on how a reply should sound.
The wider voice AI field has drawn heavy funding on both the model side—names such as ElevenLabs, Gradium and Fish Audio—and the infrastructure side, where firms including LiveKit and Vapi have raised multi-million-dollar rounds this year. Consolidation through partnerships and acquisitions is part of how leading players are trying to lock in position as new business models appear.
Aasan alfaaz mein
Deepgram ne Daily company kharid li hai, jo Pipecat banati hai. Pipecat ek open-source tool hai jo live awaz aur video ke saath AI models ko jodta hai. Deal ki qeemat nahi batayi gayi, lekin dono kehte hain Pipecat open source hi rahega. Jo log voice bots ya call apps banate hain, unke liye ye framework istemal karna aasaan rehne ka irada hai.
