Deepgram: Aura-2

Deepgram: Aura-2 is a text-to-speech model from Deepgram, listed on api14 under the speech category. Deepgram is the provider behind the model, and Aura-2 is built around Deepgram's canonical Aura-2 voice catalog for speech synthesis. If you already work with Deepgram's voice lineup, this is the same family of voices exposed through api14 rather than a separate or renamed variant.

The model takes text as its only input and produces speech as its only output. It is multilingual, with voices spanning English, French, Spanish, German, Italian, Dutch, and Japanese. The voice list is long and includes named options such as aura-2-thalia-en, aura-2-agathe-fr, aura-2-agustina-es, aura-2-aurelia-de, aura-2-cesare-it, aura-2-beatrix-nl, and aura-2-ama-ja, among many others. There is no context window figure published for this model, and no supported request parameters are listed, so treat it as a focused synthesis endpoint rather than a general text model.

Because the output is audio and the input is plain text, the practical uses follow directly from that shape. You can generate spoken versions of written content such as articles, documentation, or notifications. The multilingual voice set makes it suitable for products that serve several language markets from one integration, for example localised app prompts, IVR-style messages, or narrated learning material. Picking a voice per language or per persona is a matter of selecting the matching voice id from the catalog.

To call it through api14, use the OpenAI-compatible API with the model id "deepgram/aura-2". Because api14 follows the OpenAI request shape, you send your text and select a voice from the Aura-2 catalog in the request. No additional supported parameters are listed for this model, so keep the request minimal and rely on the model id plus your chosen voice. The page's live price table covers billing, so check there for current rates.

One thing to keep in mind is scope. Aura-2 only converts text to speech. It does not accept audio input, so transcription, translation of audio, or speech-to-speech work is outside what this model does. If your task is text generation, reasoning, or anything beyond synthesis, a different model on api14 will fit better. Within speech synthesis itself, the main decision is which of the many voices matches your language and tone.