ElevenLabs exited beta and launched Eleven Multilingual v2, supporting emotionally expressive speech synthesis and cross-lingual voice cloning across 28 languages while preserving a speaker's vocal timbre and cadence.
ElevenLabs
Also known as: Eleven Labs · Eleven Multilingual · Eleven Flash · ElevenLabs Scribe
Plain English
The leading AI voice and audio research lab—best known for expressive text-to-speech, instant voice cloning, automated multilingual video dubbing, sound-effect generation, and ultra-low-latency conversational voice agents.
Technical Overview
ElevenLabs develops proprietary neural audio foundation models spanning high-prosody TTS (Multilingual v2, Eleven v3), sub-100ms streaming synthesis (Flash v2.5) engineered for WebSocket voice agent harnesses, text-to-sound-effect diffusion, and high-accuracy speech recognition (Scribe) that rivals Whisper.
Timeline (4)
2023
Release2024
ReleaseElevenLabs released Flash v2.5, cutting model inference latency to ~75ms across 32 languages so real-time conversational voice agents could respond without noticeable turn-taking delay.
2025
ReleaseElevenLabs expanded from speech synthesis into automatic speech recognition with Scribe v1, achieving lower word-error rates than Whisper large-v3 across 99 languages with speaker diarization and word-level timestamps.
- Release
ElevenLabs introduced Eleven v3 in alpha (reaching general availability in February 2026), adding inline paralinguistic audio tags (such as
[whispers],[laughs], and[sighs]) and multi-speaker dialogue turn handoffs in a single generation pass.
See also (7)
Open full graphElevenLabs