ElevenLabs: Eleven Turbo v2

Eleven Turbo v2 is a text-to-speech model from ElevenLabs, distributed here under the model id elevenlabs/eleven-turbo-v2. It belongs to the speech category and is positioned by its provider as an English-only, low-latency text-to-speech system, built for developer use cases where speed matters and only English is needed. If you already know ElevenLabs as a voice vendor, this is the faster tier of their speech lineup rather than a general purpose language model.

The model takes text as its only input and returns speech as its only output. There is no context window figure published for it, so it is not a model you evaluate on token limits the way you would a chat model. The provider describes it around a 30,000-character input per request, which is the dimension that matters when you are chunking a script before synthesis. It ships with a fixed set of named voices: george, sarah, adam, alice, bella, bill, brian, callum, charlie, chris, daniel, eric, harry, jessica, laura, liam, lily, matilda, river, roger, and will. No request parameters are listed as supported for this entry.

The practical uses follow directly from low latency and English-only output. Interactive voice interfaces benefit most, since the delay between sending text and receiving audio is the thing users notice. The same applies to reading responses aloud in an assistant, generating narration for a prototype or internal demo where turnaround matters more than studio polish, and producing audio versions of short English content such as articles or notifications. Because a single request accepts up to roughly 30,000 characters, longer documents can be handled in a small number of calls rather than many short ones, which simplifies both the client code and the stitching of the resulting audio.

You call it through api14's OpenAI-compatible API using the model id elevenlabs/eleven-turbo-v2. The request shape is the one you already use for other models on the gateway: authentication via your api14 key, a JSON body carrying the model id and your prompt text, and an audio response in return. The supported parameters list for this model is empty, so plan on sending only the model id and the text, and select one of the named voices from the list above rather than assuming arbitrary voice identifiers will be accepted. If your existing code strips unknown fields before sending, there is nothing extra to configure.

Two limits are worth weighing before you commit. The model is English-only, so any non-English text will not be handled correctly, and the voice set is fixed, which rules out cloning or custom voices through this entry. Latency is the trade the provider is making, so if your priority is expressive or highly controlled delivery rather than speed, a different speech model on api14 is likely the better fit.