The latest text-to-speech models from ElevenLabs, including the low-latency v4 Turbo, are now available on the Runway and fal platforms for real-time agents and laarge-scale narration. These models allow users to clone voices with just 10 seconds of audio and use inline tags such as [whispers] or [laughs] to direct performance and soundscapes. Eleven v4 Turbo is specifically optimized for near-instant responses in over 90 languages.

Eleven v4 is built on a new architecture and currently ranks No. 1 on the Artificial Analysis Provider Voice TTS Arena with an Elo of 1,319. It also achieved a record pronunciation robustness score of 91.7%, the only model to exceed 93% in three of four measured categories. The model processes 73.4 characters per second of generation time and is priced at $80 per 1 million characters.