Voice-generating platform ElevenLabs raises $19M, launches detection software

ElevenLabs, the viral AI-powered platform for creating artificial voices, has raised a brand new spherical of money.

As we speak, the startup introduced the closure of a $19 million Sequence A spherical co-led by entrepreneurs Nat Friedman and Daniel Gross alongside Andreessen Horowitz. Different members included heavyweights Creator Ventures, SV Angel, Instagram co-founder Mike Krieger, Oculus co-founder Brendan Iribe, Deepmind and Inflection AI co-founder Mustafa Suleyman and O’Reilly Media founder Tim O’Reilly.

A supply accustomed to the matter tells TechCrunch that the tranche values ElevenLabs at $99 million post-money — a decent determine, particularly contemplating that the startup launched simply over a 12 months in the past.

“This funding might be used to proceed constructing ElevenLab’s cutting-edge analysis hub for voice AI and to launch a variety of extra merchandise to assist particular market verticals equivalent to publishing, gaming, leisure and conversational functions,” co-founder and CEO Mati Staniszewski informed TechCrunch through e-mail.

ElevenLabs, which has made headlines over the previous few months for causes each good and abhorrent, was based by Staniszewski, who beforehand labored at Palantir, and his childhood good friend Piotr Dabkowski, an ex-Google worker. Impressed by the mediocre dubbing of American films they watched rising up in Poland, their native nation, the pair set about designing a platform that might do higher — leveraging AI, in fact.

ElevenLabs can flip textual content into speech utilizing artificial voices, cloned voices or solely novel “synthetic” voices that mimic the sounds of individuals of varied genders, ages and ethnicities. The corporate’s AI text-to-speech fashions are language-agnostic, permitting company prospects to fine-tune them and construct their very own, proprietary speech fashions on high.

Coinciding with the Sequence A elevate, 15-employee ElevenLabs is launching Tasks, a workflow for modifying and creating long-form spoken content material. With Tasks, customers can generate dialogue segments and even audiobooks with out having to go away the platform.

“For business-to-business companions, our know-how can be utilized in areas equivalent to scalable and multilingual audiobook creation, voicing characters in video video games, voicing digital articles, supporting the visually impaired to entry on-line written content material and powering AI radio,” Staniszewski stated.

ElevenLabs, which launched in beta in late January, picked up steam moderately shortly — owing to the extraordinarily top quality of its generated voices, speedy technology occasions and beneficiant free tier. However as alluded to earlier, the publicity hasn’t all the time been optimistic — notably as soon as unhealthy actors started to use the platform for their very own ends.

ElevenLabs gives instruments to clone — or generate from scratch — realistic-sounding voices, leveraging AI.

4chan, the notorious message board recognized for its conspiratorial content material, used ElevenLabs’ software to share hateful messages mimicking celebrities just like the actor Emma Watson. Elsewhere, The.Verge’s James Vincent was capable of faucet ElevenLabs to clone targets’ voices in a matter of seconds — producing audio samples containing every thing from threats of violence to expressions of racism and transphobia.

In response, ElevenLabs stated that it might introduce a set of recent safeguards, like limiting voice cloning to paid accounts, banning customers who repeatedly violate its phrases of service and offering a brand new AI detection software.

The detection software launches right this moment. Known as AI Speech Classifier and obtainable as an API to “chosen” companions, it’s designed to detect whether or not an uploaded audio pattern accommodates AI-generated content material from ElevenLabs.

“Guaranteeing Generative AI platforms might be embraced safely is a key problem for the entire AI-generated sector, together with textual content, picture and voice platforms,” Staniszewski stated. “We should make sure that persons are educated concerning the nature of the generative media panorama and know that such content material is on the market — we’re dedicated to constructing instruments to assist individuals detect AI-generated content material, within the curiosity of transparency.”

A voluntary detection software — assuming it even works as marketed — received’t essentially deter unhealthy habits. However there’s one other elephant within the room that ElevenLabs hasn’t addressed: the existential menace its tech poses to voice actors.

Motherboard writes about how voice actors are more and more being requested to signal rights to their voices away in order that shoppers can use AI to generate artificial variations that might ultimately substitute them — generally with out extra compensation. Inside emails seen by The New York Occasions, meanwile, point out that Activision Blizzard, one of many greatest recreation publishers on this planet, is engaged on instruments for AI-assisted “voice cloning.”

It might seem that ElevenLabs sees this because the pure development of issues, touting its work with publishers like Storytel and media platforms like TheSoul Publishing and MNTN for audiobooks, video video games and radio content material. (Storytel and TheSoul Publishing are strategic traders.) The corporate claims that it has over one million registered customers throughout the inventive, leisure and publishing areas who’ve created ten years’ price of audio content material.

ElevenLabs plans to ultimately prolong its AI fashions to voice dubbing, following within the footsteps of startups like Papercup and Deepdub and constructing what it calls “a basis to have the ability to switch feelings and intonation from one language to a different.”

“This can allow any video to be dubbed into any language in an attractive, efficient, and scalable means, all whereas sustaining the unique speaker’s voice,” ElevenLabs writes in a press launch. “[We are] already conducting a variety of assessments with business companions to allow AI dubbing at scale.”

With $21 million within the financial institution ($2 million of which got here from a pre-seed spherical in January), ElevenLabs — penalties be damned — is laser-focused on beating again its rivals within the burgeoning generative voice area. They embrace incumbents like Amazon, Google and Microsoft in addition to startups like Murf, Tavus, Resemble AI, Respeecher, Play.ht and Lovo.