Deepgram to buy Daily, combining speech models with Pipecat voice framework

Deepgram is buying Daily, the company behind the Pipecat open-source voice agent framework. The company says the combination will give developers a complete voice AI stack, and that Pipecat stays open and vendor-neutral. The announcement, dated October 29, 2026, pairs Deepgram's speech models with Pipecat's orchestration layer and Daily's real-time audio transport, which the company describes as a global network of 75 points of presence built for live AI latency and reliability.

Deepgram argues that the pieces of a voice agent have to work as one system. The company says Pipecat will remain under its existing BSD V2 license and that Deepgram will invest in its development. The companies are establishing an independent Pipecat Foundation alongside many of the organizations that contribute to and use the framework. The stated goal is that developers can keep choosing models, tools, network transport and APIs from different providers.

Kwindla Hultman Kramer, co-founder of Daily, will become Deepgram's Head of Open Source. He said keeping Pipecat open was fundamental to his decision to join. Deepgram CEO Scott Stephenson said acquiring Daily would let the company tackle the whole voice system and offer developers a faster path from prototype to production.

Alongside the acquisition, Deepgram is introducing the Real-Time Context Standard, an open protocol co-developed with the Pipecat team. The company argues that words pass from one model to the next in a voice pipeline while surrounding audio information, such as tone, who is speaking and background sound, gets lost. The standard defines three kinds of metadata. The first is context taken from the audio, such as a whisper or an unclear word. The second is instructions for how a language model should use those signals. The third is guidance on how the reply should sound. Adapters translate these signals for each model, which is how the company says developers can use the same approach with models from different providers.

The announcement also previews a forthcoming Deepgram Voice Agent Platform that combines Deepgram's models with Pipecat. It adds two new conversational models, expanded language support and a Global API meant to improve connection reliability and cut latency.