ElevenLabs: Eleven v4

ElevenLabs: Eleven v4 is a text-to-speech model from ElevenLabs, released by the provider on 7 October 2026. It sits in the speech category and is described by ElevenLabs as its most expressive model. If you are building anything that needs to turn written text into spoken audio, this is the model to evaluate first from the ElevenLabs line.

The model takes text as its only input and returns speech as its only output. It supports inline audio tags, which let you steer emotional tone and delivery directly inside the text you send rather than through separate controls. ElevenLabs states support for more than 90 languages, so the same model can cover multilingual products without swapping to a different system per locale. The provider description also mentions a 10,000-character input limit. No context length in tokens is published for this model, and no request parameters are listed as supported, so plan around the character limit and the text in, speech out shape of the model.

The practical uses follow from those traits. Inline audio tags make it a fit for narration where you need a line delivered with a specific emotion, such as audiobooks, character dialogue in games, or ads that need a particular read. The 90+ language coverage suits localization pipelines, e-learning courses, and voiceovers for video at scale, where you would otherwise book speakers per language. Because output is speech, it also works for accessibility features, reading articles or documents aloud, and interactive voice experiences where responses are generated from text at request time.

You call it through api14's OpenAI-compatible API using the model id "elevenlabs/eleven-v4". The request follows the same shape as other OpenAI-compatible speech calls: send your text and the model returns audio. A set of named voices is available, including george, sarah, adam, alice, bella, bill, brian, callum, charlie, chris, daniel, eric, harry, jessica, laura, liam, lily, matilda, river, roger, and will. Pick the voice that matches your product's tone, and use inline audio tags in the text for emotional and delivery control. No additional request parameters are listed as supported for this model, so keep requests to the documented fields.

Keep the 10,000-character input limit in mind. For long documents you will need to split the text into chunks and stitch the resulting audio together, and for very short prompts the expressiveness may be more than you need. If your task is not speech generation, a text or multimodal model will fit better, since Eleven v4 only accepts text and only produces speech.