\n<\/p>\n
ElevenLabs<\/a>, one of the more popular startups working in the field AI audio, said Thursday that it has raised a Series C round of $180 million, valuing the company at $3.3 billion post-money. a16z and ICONIQ Growth are co-leading investment. <\/p>\n Rumors of the fundraise were first reported by TechCrunch<\/a>. The final numbers confirm some but not all of the details we previously reported (specifically, the overall size of the round is smaller than we had heard; the valuation and lead investors are the same).<\/p>\n The funding will be used to continue building out ElevenLabs\u2019 audio tools and for business development. <\/p>\n Mati Staniszewski, the CEO who co-founded the company with childhood friend Piotr Dabkowski, said in an interview that the startup is focusing its research on building audio AI models that are more expressive and have more control. Staniszewski added that the company is also focusing on \u201comni-models\u201d: combining text-based models with its audio models for multimodal interactions.<\/p>\n There has been a frenzy of investor interest in ElevenLabs going back several months<\/a>, on the back of two important currents. First, there has been a huge wave of hype around generative AI that has been catching a lot of companies in its wake. Second, ElevenLabs has emerged as a major player among those providing synthetic voice technology. Dozens of major publishers and content creators across verticals like media and gaming, as well as a number of other tech startups, are all using ElevenLabs\u2019 technology to power their voice and audio features.<\/p>\n Unsurprisingly, that has translated into a very crowded funding round with a lot of prominent names. <\/p>\n New investors in this Series C include NEA, World Innovation Lab (WiL), Valor, Endeavor Catalyst Fund, and Abu Dhabi investment firm Lunate. Past investors also participating include Sequoia Capital, Salesforce Ventures, Smash Capital, SV Angel, NFDG, and BroadLight Capital.<\/p>\n Alongside these, ElevenLabs is also picking up a number of new strategic backers \u2014 that is, companies using its technology who are now investing in it, too. These include Deutsche Telekom, LG Technology Ventures, HubSpot Ventures, NTT DOCOMO Ventures, and RingCentral Ventures.<\/p>\n ICONIQ partner Seth Pierrepont will join the company\u2019s board, alongside existing board members Jennifer Li from a16z and the co-founders of the company.<\/p>\n ICONIQ has been ramping up its activities around generative AI startups. Tapping into written output, the firm also co-led a $200 million round in Writer<\/a> last November. <\/p>\n \u201cWe have always felt that audio is a very important modality, and we thought there will be a very big company built in this category,\u201d Pierrepont told TechCrunch. \u201cWe have observed ElevenLabs from its launch, and we were impressed by the quality of the technology, how quickly it ascended in terms of mindshare and momentum, and the depth of domain expertise of the founders.\u201d<\/p>\n Pierrepont added that as a board member, a lot of the conversations with the company will be around creating new use cases for audio and finding the right markets for it. <\/p>\n At a time when startups are still finding it challenging to close growth rounds, it\u2019s notable that ElevenLabs raised its Series B round of $80 million, which valued it at $1 billion, just a year ago<\/a>. ElevenLabs has raised a total of $281 million to date.<\/p>\n In addition to a focus on improving its AI models, the company plans to use the funding to grow its conversational AI builder with an ambition to reach more consumers directly and through partnerships.<\/p>\n Last year, the company debuted an AI conversational agent platform<\/a>, and a key part of that product was developing a speech-to-text component. Staniszewski noted that the company wants to improve in that area a lot more.<\/p>\n \u201cWe want to understand what\u2019s being said by you in a conversation better. We are working on ways to move away from only generating content and understanding and transcribing speech,\u201d Staniszewski said. \u201cMany people say that speech-to-text is a solved problem. But for many languages, it is pretty bad. We think we can build better speech detection models because we have in-house teams to annotate data and give us quick feedback.\u201d<\/p>\n The company also wants to double down on creating AI-powered conversational agents by supporting legacy communications like telephony and better integrating different kinds of knowledge sources. This is partly why it is partnering with telcos in this round.<\/p>\n It\u2019s also being used by its customers to tap into their own archives. Last year, ElevenLabs partnered with TIME publication to deploy a conversational bot<\/a> for users to ask questions about TIME Person of the Year.<\/p>\n Staniszewski said the company envisions more conversational AI agents on sites: on news sites, for example, users would be able to ask questions about stories or ask the bot to summarize them.<\/p>\n The CEO also noted that while AI-powered voice bots\u2019 quality has improved, the problem of sounding natural while reacting to humans speaking or emoting in different ways has not been solved yet.<\/p>\n \u201cThe way I speak to you impacts how you react or respond to me. Sometimes, I\u2019ll be excited, or sometimes, I\u2019ll be calm, and at times, I will interrupt you. You will respond to me accordingly. Current-gen AI solutions are on the verge of being good, but they are not as good as humans,\u201d Staniszewski said.<\/p>\n ICONIQ\u2019s Pierrepont also emphasized that if AI doesn\u2019t understand you well when you are talking, machine communication breaks down and users immediately lose interest.<\/p>\n ElevenLabs has mostly grown its reach (and revenue funnel) by way of B2B partnerships. But it\u2019s also going out on a direct limb, too. <\/p>\n In 2024, the startup launched its first purely consumer-facing product, ElevenLabs Reader, an app that reads out articles, text, and documents<\/a>. Later, the company added the ability to create a podcast with generative AI voices<\/a> from documents and web pages \u2014 not unlike what you can do with Google\u2019s NotebookLM. Staniszewski said that it wants to expand into more consumer experiences.<\/p>\n It may actually already be doing that. TechCrunch spotted that the company has been testing a program on the ElevenLabs Reader app inviting users to publish audiobooks on the platform. The company also wants to give tools to creators to have multiple voices read out an audiobook in the future while also creating better localization.<\/p>\n Staniszewski noted that the company is figuring out ways for users and companies to better distribute their content, including on its own app. Whether that brings it into actual direct competition with its customers will be something to watch. (That has been one reason why many B2B tech companies prefer to stay away from direct-to-consumer plays.) Notably, ElevenLabs powers voice technology for audio content platforms like Lightspeed-backed Pocket FM<\/a> and Google-backed Kuku FM<\/a>.<\/p>\n ElevenLabs already powers AI-generated audio on products and platforms like Perplexity, Rabbit R1, Chess.com, ESPN, Lex Fridman podcast, The Atlantic, and Synthesia. The goal for the company is to be in more places and also own an end-to-end conversation stack so it can generate more experiences and insights for its customers.<\/p>\n Not all of ElevenLabs\u2019 silver linings have been without clouds: its tech has been implicated in a few notable misinformation campaigns. A recent report from threat intelligence company Recorded Future found that the company\u2019s product was used in a Russian propaganda operation<\/a>. Last year, someone used the company\u2019s voice platform to create an audio deepfake of Joe Biden<\/a>. In 2023, Motherboard<\/a> reported that 4chan members allegedly used the AI audio generation tool to create voices that sounded like Joe Rogan, Ben Shapiro, and Emma Watson to spread problematic content.<\/p>\n But the company has been quick to respond. Today, it has a policy prohibiting \u201cunauthorized, harmful, or deceptive impersonation.\u201d Plus, it uses a mix of machine-led and human moderation to weed out such content. However, as the company grows its set of tools and has more direct consumer touchpoints, this opens the door to more opportunities for malicious actors to look for ways to misuse it.<\/p>\n \u201cAs one of the frontrunners of AI audio work, we do treat it as our responsibility to build the right safety mechanism as we build out the technology. We will frequently make choices to prioritize safety over speed of deployment or commercial benefit,\u201d Staniszewski said.<\/p>\n Staniszewski added that while the company follows C2PA, a standard to track content using metadata, it also has a public tool<\/a> that allows anyone to check if audio was generated through ElevenLabs technology using digital signatures it places in the audio during generation. That could also be a track that continues to develop over time as approaches for misuse also become more sophisticated.<\/p>\n<\/div>\nThe product roadmap<\/h2>\n
Safety<\/h2>\n