EElevenLabs

Eleven v3 Conversational is ElevenLabs' expressive realtime speech model. The model ID is eleven_v3_conversational. ElevenLabs documents about 280ms model latency, excluding application and network latency, and 70+ languages. It is the model the docs recommend for support agents, assistants, and interactive characters, including the Text to Dialogue WebSocket.

API ID eleven_v3_conversational

Eleven v3 is ElevenLabs' documented flagship for human-like, expressive speech. The model ID is eleven_v3. It supports 70+ languages and is used with the Text to Speech API and the Text to Dialogue API. The official models page lists a 5,000-character request limit, about five minutes of audio. Help Center still calls Eleven v3 the latest speech model; Eleven v4 is not on the models page.

API ID eleven_v3

Eleven Flash v2.5 is ElevenLabs' ultra-fast speech model for realtime use. The model ID is eleven_flash_v2_5. ElevenLabs documents about 75ms model latency and support for the Multilingual v2 languages plus Hungarian, Norwegian, and Vietnamese. The models page lists a 40,000-character request limit, about 40 minutes. Number normalization is off by default to keep latency low.

API ID eleven_flash_v2_5

Eleven Music v2 is ElevenLabs' current music model. The model ID is music_v2. It generates studio-grade music from text prompts, composition plans, and previously generated songs, in English, Spanish, German, Japanese, and more. ElevenLabs says music_v1 is outclassed by music_v2.

API ID music_v2

Scribe v2 is ElevenLabs' documented state-of-the-art speech recognition model. The model ID is scribe_v2. It supports 90+ languages. The realtime sibling on the same models page is scribe_v2_realtime. Scribe v1 is listed as outclassed by the v2 models.

API ID scribe_v2