ElevenLabs Turbo v2.5

Turbo v2.5 is optimized for sub-300ms response times, making it suitable for real-time conversational applications such as voice assistants and interactive AI agents. This low-latency performance is a key advantage over earlier ElevenLabs models that prioritized output quality over speed.

Turbo v2.5 is a multilingual text-to-speech model, supporting multiple languages for voice synthesis. The model aligns with ElevenLabs' broader multilingual platform, which covers dozens of languages with a range of synthetic voices.

Turbo v2.5 is specifically built for real-time conversational applications where latency is critical, such as AI phone agents, live voice chat, and interactive game NPCs. ElevenLabs also offers higher-quality models better suited for offline content production like audiobooks or video dubbing where sub-second response time is not essential.

Turbo v2.5 builds on the original Turbo model with improvements to voice quality and language coverage while maintaining the low-latency optimization that defines the Turbo line. The v2.5 designation, released in 2024, indicates an incremental upgrade rather than a fundamentally new architecture.