Audio
ElevenLabs Turbo 2.5
Use it ↗ElevenLabs Turbo 2.5 offers low-latency text-to-speech in 32 languages with fast, cost-effective generation, starting from 8 CUs.
Released in late 2025, ElevenLabs Turbo 2.5 is a text-to-speech model that prioritizes speed and cost-efficiency for real-time applications. It is well-suited for conversational AI, interactive games, and high-volume projects where low latency is a primary requirement. The model supports 32 languages and generates speech about three times faster than Multilingual v2, with a per-character cost that is 50% lower. This speed comes at the cost of the emotional nuance found in the higher-fidelity Multilingual v2.