Voice AI company Deepgram said on Oct. 29 that it’s acquiring Daily, the company behind Pipecat, one of the most widely used open-source frameworks for building voice agents. Neither company disclosed the price.
A voice agent is software that holds a spoken conversation, usually by chaining a speech-to-text model, a language model and a text-to-speech model. Pipecat manages the audio and video streams that pass through that chain and tries to cut the delay between each step, TechCrunch’s Ivan Mehta explained in his report on the deal.
Daily also runs a hosted real-time network for voice and video calls. Deepgram says Pipecat and that network will sit underneath its new Voice Agent Platform, alongside Deepgram’s own speech models.
What happens to Pipecat
Deepgram says Pipecat will stay free and open source under its existing BSD license, a permissive license that lets anyone use and modify the code. The companies are also setting up an independent Pipecat Foundation with organizations that contribute to and use the framework, and Deepgram says Pipecat will stay vendor neutral, so developers can mix models and services from different providers.
“Keeping Pipecat open was fundamental to our decision to join Deepgram,” Daily co-founder Kwindla Hultman Kramer said in the announcement. He joins Deepgram as head of open source, and the rest of Daily’s team moves over too.
Kramer told TechCrunch that several companies had shown interest in buying Daily. He said he chose Deepgram because CEO Scott Stephenson was “super committed to Pipecat being open source and vendor-neutral forever,” and he called openness a commercial advantage for a company that sits in the middle of a fast-growing ecosystem.
Stephenson told TechCrunch that he wants developers to use whatever models fit, including competitors’. “It’s great if you’re using Deepgram models, but I want you to be using the best models for your use case,” he said.
The two companies go back a long way. TechCrunch reported that both were in the same Y Combinator batch in 2016. Daily has raised more than $62 million from investors including Tiger Global and Nat Friedman, according to PitchBook data cited by TechCrunch, and Deepgram reached a valuation of more than $1 billion earlier this year with a $130 million Series C.
Kramer told TechCrunch that Daily started experimenting with voice conversations between people and AI around the time GPT-4 came out, work that grew into Pipecat.
A new protocol for context
Deepgram held its first Deepgram Speak event in San Francisco the same day. There, Stephenson introduced the Real-Time Context Standard, an open protocol built with the Pipecat team.
Deepgram says the protocol fills a gap in today’s voice systems. Words get passed from one model to the next, but signals around them, such as tone, who’s speaking or background noise, get lost. The standard adds metadata for those audio cues and for how the reply should sound, so an agent can ask someone to repeat an unclear word or answer quietly when a person is whispering.
“Acquiring Daily allows us to bring that expertise together to tackle the whole voice system and give developers a faster path from prototype to production,” Stephenson said in the announcement.
Source: Deepgram, “Deepgram is acquiring Daily, makers of Pipecat”
Additional sources: TechCrunch, Ivan Mehta, “Deepgram acquires Pipecat-maker Daily to bolster its voice AI infrastructure”; Daily blog, Daily, “Deepgram Is Acquiring Daily and Establishing an Independent Pipecat Foundation”
