ElevenLabs Review

For twenty years text to speech sounded like text to speech, and ElevenLabs is the product that changed that. Its voices carry pacing, emphasis and breath in a way that fools most listeners on a first hearing, and the multilingual model holds a speaker's identity across languages, which makes the dubbing feature the most impressive thing it does. Voice cloning from a minute of audio is accurate enough to be alarming, and the company has had to build consent verification around it for that reason.

For the repetitive work the tool is designed for, narration at scale, audiobooks, accessibility, localisation, nothing else is close. The limits are the ones of the medium: it reads what it is given and cannot judge what it is reading, so a badly written script is delivered beautifully, and long passages occasionally drift in tone without a human ear to catch it. The pricing climbs steeply with character volume. It is the category leader by a wide margin and knows it.