Deepgram acquires Pipecat-maker Daily to bolster its voice AI infrastructure

The acquisition integrates Pipecat, a framework that manages real-time voice and video streams alongside components like speech-to-text (STT), large language models (LLMs), and text-to-speech (TTS) models, into Deepgram's offerings. This system is designed to reduce latency in AI-driven communication pipelines. Daily, Pipecat's creator, also provided hosted infrastructure for various voice and video calls.

Deepgram's CEO, Scott Stephenson, indicated the company's long-standing interest in open source, identifying an orchestration framework for agents as a critical area. By keeping Pipecat open source, Deepgram aims to foster a wider ecosystem for voice agent development, encouraging developers to build with Pipecat even if they use other models alongside Deepgram's. This approach, as noted by Daily's co-founder, Kwindla Kramer, can be a "commercial advantage" by positioning the framework centrally within a growing ecosystem.

The deal allows Deepgram to explore applications beyond just voice, as Pipecat supports real-time communication in various formats. Deepgram is also launching the Real-Time Context Standard (RCS), an open protocol co-developed with the Pipecat team, which defines metadata for voice agents, including context from audio and instructions for language models.