Deepgram收购Daily：开源框架Pipecat保持免费、两家公司同出2016年Y Combinator
语音SI基础设施公司 Deepgram 10月29日在旧金山（San Francisco）举行的首届 Deepgram Speak 大会上宣布，正在收购实时音视频基础设施公司 Daily。Daily 是开源语音智能体框架 Pipecat 的创建者，Deepgram 称 Pipecat 是目前全球最受欢迎的开源语音智能体框架。两家公司没有披露收购金额。
Deepgram 表示，Pipecat 与 Daily 的全球实时网络将与自家语音模型一起，支撑其新推出的 Voice Agent Platform（语音智能体平台），组成其所称首个完整、开放、可用于生产环境的对话式语音SI平台，开发者可以在保留架构与模型选择权的同时构建和扩展智能体。Daily 的全球网络共有75个接入点，专为实时SI的低延迟与高可靠性需求设计。
收购完成后，Daily 团队将整体加入 Deepgram。Daily 联合创始人 Kwindla Hultman Kramer 将出任 Deepgram 的开源负责人。他表示，保持 Pipecat 开源是 Daily 决定加入 Deepgram 的根本原因。Deepgram 联合创始人兼首席执行官 Scott Stephenson 表示，Kramer 与 Daily 团队深谙如何让实时对话可靠运行，这次收购把双方的专长合到一起处理整个语音系统，让开发者更快从原型走到生产。
Pipecat 将继续完全开源，在现行 BSD 许可下对所有开发者免费。两家公司还将与众多参与和采用 Pipecat 的机构共同设立一个独立的 Pipecat Foundation（基金会），Deepgram 承诺持续投入其开发与社区，并保持开源与厂商中立，开发者可以自由组合不同厂商的模型、工具、网络传输与接口。Deepgram 还将继续销售、支持和开发 Daily 现有的全部产品与服务，包括 WebRTC、电话与跨平台库。
同场，Stephenson 还推出了与 Pipecat 团队共同开发的开放协议 Real-Time Context Standard（实时上下文标准，RCS）。Deepgram 称，现有语音系统在模型之间传递的往往只是文字，语气、说话人身份与背景状况等丰富的音频上下文会丢失；RCS 定义了三类新元数据——来自音频的上下文、语言模型如何使用这些信号的指令、以及回复应当如何发声的指引——让智能体能在没听清时要求澄清，或在对方耳语时放轻声音回应。
据 Deepgram 介绍，其语音模型与实时基础设施已被全球逾5,000家机构用于生产环境，超过65万名开发者使用其技术处理了合计逾20万年的音频。Daily 此前累计融资超过6,200万美元，投资方包括 Tiger Global 等机构；Deepgram 今年早些时候以13亿美元的 C 轮融资达到独角兽估值。两家公司同为 Y Combinator 2016 年批次的学员公司。
Deepgram Acquires Daily: Pipecat Stays Open Source as the Two 2016 Y Combinator Companies Combine
Voice AI infrastructure company Deepgram announced on October 29, at its inaugural Deepgram Speak event in San Francisco, that it is acquiring Daily, the real-time audio and video infrastructure company behind Pipecat, which Deepgram calls the world's most popular open-source voice agent framework. The companies did not disclose the price.
Deepgram said Pipecat and Daily's global real-time network will underpin its new Voice Agent Platform alongside its own speech models, forming what it described as the first complete, open, production-grade platform for conversational voice AI, letting developers build and scale agents while keeping control over their architecture and model choice. Daily's global network has 75 points of presence engineered for the latency and reliability demands of live AI.
Following the acquisition, the Daily team will join Deepgram. Daily co-founder Kwindla Hultman Kramer joins as Head of Open Source. 'Keeping Pipecat open was fundamental to our decision to join Deepgram,' Kramer said. Deepgram CEO and co-founder Scott Stephenson said Kramer and the Daily team deeply understand what it takes to keep a live conversation working reliably, and that acquiring Daily brings that expertise together to tackle the whole voice system and give developers a faster path from prototype to production.
Pipecat will remain fully open source and free to all developers under its existing BSD license. Together with many of the organizations that have contributed to and use Pipecat, the companies are establishing an independent Pipecat Foundation, and Deepgram said it will invest in Pipecat's continued development and community while keeping it open source and vendor neutral, so developers can combine models, tools, network transport and APIs from different providers. Deepgram will also sell, support and develop all of Daily's current products and services, including WebRTC, telephony and cross-platform libraries.
At the same event, Stephenson introduced the Real-Time Context Standard (RCS), an open protocol co-developed with the Pipecat team. Deepgram said today's voice systems pass words from one model to the next but lose the rich audio context around them, such as tone, who is speaking and what is happening in the background. RCS defines three new types of metadata — context from the audio, instructions for how a language model should use those signals, and guidance for how the response should sound — so an agent can ask for clarification when it did not hear clearly or respond quietly to someone whispering.
Deepgram said its speech models and real-time infrastructure power conversational voice AI in production for more than 5,000 organizations globally, and that more than 650,000 developers have used its technology to process more than 200,000 years of audio. Daily had raised more than $62 million from investors including Tiger Global, according to PitchBook data cited by TechCrunch, and Deepgram reached unicorn status earlier this year with a $130 million Series C round. Both companies were in the same Y Combinator batch in 2016.
来源：GSM·美国（原文链接）
图片来源：图：San Francisco 市中心航拍（资料图），Deepgram 10月29日在该市首届 Speak 大会宣布收购 Daily。照片：Spicypepper999，CC0，经 Wikimedia Commons。 / Downtown San Francisco from the air (file photo). Photo by Spicypepper999, CC0, via Wikimedia Commons.
