ElevenLabs, whose voice models are used for synthetic narration, dubbing and music, is worth roughly $22 billion in the eyes of its investors, according to TechCrunch reporting on 24 September that StartupFortune cites. That is roughly double the $11 billion set in its Series D in February, and follows a July report from Bloomberg that the company was weighing an employee share sale at about that level. ElevenLabs, founded by Polish entrepreneurs four years ago, says it is heading for $600 million in annual recurring revenue, more than 55% of it from business customers such as Klarna and Cisco.
Speaking to TechCrunch at the Nrth conference in Toronto, chief executive Mati Staniszewski said he is willing to accept lower gross margins to keep gaining market share. He was also pressed on when people ought to be told that a voice they hear is synthetic, a question StartupFortune says still has no settled answer.
Part of the pressure comes from free models. In March Mistral released Voxtral TTS, a 4-billion-parameter text-to-speech model whose weights are available under a non-commercial licence and which can run on a single consumer graphics card. Mistral's own listening tests had it beating ElevenLabs' Flash v2.5 in most comparisons, though it handles nine languages to ElevenLabs' 70-plus.
ElevenLabs has leaned on enterprise channels in response, building its speech tools into IBM's watsonx Orchestrate in March with more than 10,000 voices and compliance features for payments and health data. It has faced misuse before: after its tools were tied to the 2024 robocall imitating Joe Biden in New Hampshire, it banned the account involved and cited protections including a classifier that detects its own speech and a bar on cloning prominent public figures.
Sources · 1