Now hear this: voice cloning AI startup ElevenLabs nabs $19M

ElevenLabs, a year-old AI startup from former Google and Palantir workers that’s targeted on creating new text-to-speech and voice cloning instruments, has raised $19 million in a Sequence A spherical co-led by Andreessen Horowitz (a16z), former Github CEO Nat Friedman and former Apple AI chief Daniel Gross, with further participation from Credo Ventures, Idea Ventures, and an array of strategic angel buyers together with Instagram’s co-founder Mike Krieger and Oculus VR co-founder Brendan Iribe together with many others.

As well as, Andreessen Horowitz is becoming a member of ElevenLabs’ board, citing the late Martin Luther King Jr.’s “I Have a Dream” speech in its weblog submit on the information, as one of many examples of how the human “voice carries not solely our concepts, but additionally essentially the most profound feelings and connections.”

ElevenLabs was based by Piotr Dabkowski, an ex-Google machine studying engineer, and Mati Staniszewski, an ex-Palantir deployment strategist, to develop ultra-realistic text-to-speech fashions for training, audiobooks, gaming, films, enterprise, and extra.

Each grew up in Poland and had been impressed to create the corporate after watching poorly dubbed movies from the U.S., in response to a16z.

“Think about the probabilities unlocked when creatives can provide any character any voice, and have that voice be indistinguishable from the unique,” a16z writes in its weblog submit saying its participation within the spherical.

At present out there merchandise from ElevenLabs embody its Speech Synthesis, which converts any writing to “skilled audio, quick”; VoiceLab, which permits clients to clone voices from just some brief snippets of audio and may create completely new artificial voices; and the newly unveiled AI Speech Classifier with an API, which permits “anybody to add an audio pattern and discover out whether or not the clip incorporates AI-generated audio from ElevenLabs,” which the corporate says is a first-of-its-kind software.

Up subsequent from ElevenLabs are two different deliberate merchandise: an “AI dubbing software” that may enable customers to take present recordings of speech and convert it into one other language whereas preserving the unique voices, and Publishers Tasks, an audio manufacturing app to regulate a audio system’ pacing, insert pauses, and assign explicit audio system to completely different textual content fragments.

Already, ElevenLabs claims to have over one million creators and builders from varied industries, who’ve collectively generated greater than 10 years of audio content material. Amongst its early clients embody audiobook writer Storytel and sport developer Paradox Interactive, in addition to a brand new AI-focused podcast from entrepreneur Seth Godin.

“Our mission is to dissolve language limitations and put all audiences inside attain of content material creators in a protected and accountable manner,” mentioned Staniszewski.

Acknowledging that there are critical dangers of hurt with voice-cloning tech particularly, ElevenLabs has posted a “Voice Cloning Information” on its web site that states: “you can’t clone a voice for abusive functions resembling fraud, discrimination, hate speech or for any type of on-line abuse with out infringing the regulation.” Nevertheless, “caricature, parody and satire” and “creative and political speech contributing to public debates” are makes use of that the corporate permits and helps.

Whereas ElevenLabs is cautious to state its on-line information “doesn’t represent authorized recommendation,” it does word that it might probably droop the accounts and content material of any customers who it deems in violation, at will, and that “we might also report any criminality in accordance with relevant legal guidelines to the authorities or work with authorities on additional motion.”

The information of the funding spherical and launch of recent instruments comes simply days after Meta Platforms’ printed a analysis paper describing its personal generative AI voice synthesis software Voicebox, although Meta has but to publicly launch the software to customers out of issues for abuse.