Deepgram is acquiring Daily, bringing the team behind Pipecat into a company known for voice AI models and infrastructure. The companies have not disclosed the deal amount and say Pipecat will remain open-source.
What Pipecat brings to Deepgram
Pipecat coordinates the components involved in real-time voice and video applications. It manages streams and connects pieces such as speech-to-text, large language models and text-to-speech, while reducing latency.
Daily also offers hosted infrastructure for voice and video calls. Together, those capabilities give Deepgram a closer view of how real-time systems are assembled and how different models and services work together.
Deepgram says the acquisition will help developers build agents with expressive, end-to-end speech systems. They can connect different models and use a context bus to link parts of the system. The framework also supports real-time communication in formats beyond voice, opening a path for Deepgram to explore other areas.
Keeping the framework open
The companies have not disclosed financial terms. They have said Pipecat will stay open-source, a commitment that Daily co-founder Kwindla Kramer said helped make Deepgram the right match.
Kramer said Daily began experimenting with AI-to-human communication over voice and text around the time GPT-4 came out. That work led to Pipecat. He said he chose Deepgram partly because CEO Scott Stephenson was committed to keeping the framework open-source and vendor-neutral, and because the engineering cultures matched.
Stephenson described openness as part of Deepgram’s plan to help developers build voice agents using the models best suited to their needs. That could mean Deepgram models or models from partners. He said feedback from that broader ecosystem can also show Deepgram where its own models and products need to improve.
Daily had raised over $62 million in funding from investors including Tiger Global, Highfield Capital, Nat Friedman, Lachy Groom and Kevin Weil, according to PitchBook data. Deepgram reached unicorn status earlier this year after a $130 million Series C round. The two companies were in the same Y Combinator batch in 2016.
A shared protocol for voice agents
Alongside the acquisition, Deepgram is launching the Real-Time Context Standard, or RCS, an open protocol it co-developed with the Pipecat team. It sets out metadata that voice agents can use, including context from audio such as a whisper.
RCS also describes instructions for how a language model can use those signals and guidance for how its response should sound. In practical terms, the protocol is intended to help the parts of a voice agent share context and coordinate their behavior.
Kramer said Daily did not view Pipecat as only a voice toolkit. The team saw it as a toolkit for a new kind of AI-native software, built around real-time communication. That framing helps explain why the acquisition may matter beyond Deepgram’s core voice offerings.
Competition across the voice AI stack
The deal comes as investment is flowing into both voice AI models and the infrastructure around them. Model makers including ElevenLabs, Gradium and Fish Audio have raised millions of dollars. Infrastructure companies such as LiveKit and Vapi have also raised multi-million-dollar funding rounds this year.
Deepgram’s acquisition combines model expertise with a framework and hosted services used to build real-time applications. Keeping Pipecat open-source and vendor-neutral could let developers combine components from multiple providers, while giving Deepgram a place to participate in that ecosystem and learn where its own products can improve.