语音AI部门热门，当更多的商业模式通过伙伴关系和收购而浮现时，顶级玩家正试图巩固自己的地位。 Deepgram是进行这种动作的最新方法：称公司于星期四收购了Daily，即最受欢迎的开源语音AI基础设施之一Pipecat的制造者和维护者。
公司并没有透露交易金额
对于未启动者来说，Pipecat是一个框架，它管理语音和视频流以及管道中的其他片段，如语音转文本(STT)，大型语言模型，以及文本转语音(TTS)模型，同时减少闲置。 其制作者Daily为各种语音和视频电话提供托管基础设施。
有了Pipecat,Deepgram很可能获得对实时语音和视频基础设施的更深的理解。
根据PitchBook的数据，Daily拥有超过6200万美元的融资资金，至今来自投资方，如Tiger Global,Highfield Capital,Nat Friedman,Lachy Groom和Kevin Weil. 今年早些时候，Deepgram以1.3亿美元的C系列回合达到独角兽地位。 巧合的是，这两家公司在2016年都属于同一批Y Combinator.
Kwindla Kramer,Daily联合创始人，大约在GPT-4出来时，公司开始在语音和文本上进行AI对人通信的实验，最终形成了Pipecat框架。 他注意到，虽然有很多收购利益，但他喜欢Deepgram的CEO斯科特·斯蒂芬森(Scott Stephenson)承诺永远保留Pipecat 开源，两组织的工程文化也相匹配，使得Daily的团队可以自由地在一个更大的组织内在堆栈的不同部分工作。
“Scott超级致力于Pipecat永远是开源和供应商中立。 我认为那里既有个人承诺，也有商业驱动力。 如果你坐落在一个正在快速成长的生态系统中间，并且拥有这个生态系统的许多部位
Stephenson 称认为Deepgram长期想在开源工作，公司意识到agents的管弦乐框架是一个潜在的领域，这促使决定与Daily合作。
“我们希望开放生态系统，如果你使用Deepgram模型，如果Deepgram注意到，我们在某些领域可以赶上自己的模型和产品，那就太好了。 我们希望你们能利用合作伙伴
Stephenson 称的收购让开发者通过插入不同的模型和一个连接所有部件的上下文总线来构建带有表达性和端到端语音系统的agents.
对于Deepgram来说，交易还让其通过Pipecat探索语音以外的领域，因为框架迎合实时不同格式的通信。
“我们从未将Pipecat当作一个语音工具包。 我们把它当作我们建造这种新型AI-native软件所需要的工具箱。 因为我们来自哪里，我们总是假设实时，"Kramer 称。
Deepgram也是发布实时背景标准(RCS)，是它与Pipecat团队共同开发的语音agents的开放协议。 协议定义了元数据，如音频的上下文，如whisper，语言模型如何使用这些信号的指令，以及响应应如何声音的指导。
Voice AI从模型和基础设施方面都看到了大量投资。 ElevenLabs,Gradium,Fish Audio等模型制造者拥有融资百万美元。 另一方面，LiveKit和Vapi等基础设施公司今年有几百万美元的融资资金回合。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
The voice AI sector is hot, and top players are trying to consolidate their position as more business models come to the fore through partnerships and acquisitions. Deepgram is the latest to make such moves: The company said on Thursday that it is acquiring Daily, the maker and maintainer of one of the most popular open-source voice AI infrastructures, Pipecat.
The companies didn’t disclose the deal amount but said that they will keep Pipecat open-source.
For the uninitiated, Pipecat is a framework that manages voice and video streams along with other pieces in the pipeline, like speech-to-text (STT), large language models, and text-to-speech (TTS) models while reducing latency. Its maker, Daily, provides hosted infrastructure for voice and video calls of any kind.
With Pipecat, Deepgram will likely gain a deeper understanding of real-time voice and video infrastructure.
Daily has raised over $62 million in funding to date from investors such as Tiger Global, Highfield Capital, Nat Friedman, Lachy Groom and Kevin Weil, according to PitchBook data. Earlier this year, Deepgram reached unicorn status with a $130 million Series C round. Incidentally, both companies were in the same Y Combinator batch in 2016.
Kwindla Kramer, Daily’s co-founder, said around the time when GPT-4 came out, the company started experimenting with AI-to-human communication over voice and text, which eventually resulted in the Pipecat framework. He noted that while there has been a lot of acquisition interest, he favored Deepgram’s CEO Scott Stephenson’s commitment to keeping Pipecat open-source forever, and the engineering culture of both organizations matched, giving Daily’s team the freedom to work on different parts of the stack within a bigger organization.
“Scott was super committed to Pipecat being open-source and vendor-neutral forever. I think there is both a personal commitment and commercial driver there. If you sit in the middle of an ecosystem that’s growing very fast, and have many parts of that ecosystem that are genuinely the best from your company, then everything being as open as possible is like a commercial advantage,” Kramer told TechCrunch.
Stephenson said that Deepgram has been wanting to work in open source for a long time, and the company realized that an orchestration framework for agents was a potential area, which drove the decision to partner with Daily.
“We want to open up the ecosystem so everybody can build voice agents. It’s great if you’re using Deepgram models, but I want you to be using the best models for your use case. And if Deepgram notices that we have some ground to catch up on in a certain area for our own models and products, that’s great. We want you to use partners, and then we’ll also work on becoming better in those areas,” he said.
Stephenson said the acquisition lets developers build agents with expressivity and end-to-end speech systems by plugging in different models and a context bus that connects all parts.
For Deepgram, the deal also lets it explore areas beyond voice through Pipecat, as the framework caters to real-time communication in different formats.
“We never thought about Pipecat as a voice toolkit. We thought about it as the toolkit we needed to build this new kind of AI-native software. Because of where we came from, we just always assumed real-time,” Kramer said.
Deepgram is also launching Real-Time Context Standard (RCS), an open protocol for voice agents that it co-developed with the Pipecat team. The protocol defines metadata like context from the audio, like a whisper, instructions for a language model on how it can use those signals, and guidance on how the response should sound.
Voice AI has seen a lot of investment from both the model and the infrastructure sides. Model makers like ElevenLabs, Gradium, and Fish Audio have raised millions of dollars. On the other hand, infrastructure companies like LiveKit and Vapi have raised multi-million-dollar funding rounds this year.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.