ElevenLabs launches v4 speech models with 90+ language support

ElevenLabs has launched its latest speech models, ElevenLabs v4 and v4 Turbo, featuring a new architecture designed for enhanced expression control and faster voice cloning. The company claims users can now clone a voice using only 10 seconds of audio.

The new models support over 90 languages, an increase from the 70 languages supported by the previous v3 model. Significant quality improvements were reported in Japanese, Mandarin, Cantonese, and Brazilian Portuguese. The v4 model also utilizes text context to provide more nuanced expression and allows for expanded use of inline tags to define vocal delivery.

Designed with an enterprise focus, the v4 model is optimized for voice agents by offering lower latency to facilitate more fluid conversations. The model can begin generating audio as soon as the underlying large language model starts producing answers. This update comes as ElevenLabs, which recently raised $500 million at an $11 billion valuation, faces increasing competition from companies such as OpenAI, Google, Cartesia, and Deepgram.