MenaVoice has four text to speech models and one speech to text model. The text to speech models share the same API, the same studio voices (except Clone, which speaks with your library) and the same price, so you can switch between them with a single parameter.
Text to Speech models
| Model | quality | Description | Dialect steering |
|---|---|---|---|
| MenaVoice 2Newest | v2 | Our most expressive model. Output varies between takes. | |
| MenaVoice 1.5 | pro | Richer, studio-style delivery for narration and ads. | |
| MenaVoice 1vDefault | flash | Fast and consistent for everyday scripts. | |
| MenaVoice Clone | clone | Speaks with any voice from your voice library. |
Choose a model with the quality parameter of POST /tts. Leave it out to use MenaVoice 1v.
MenaVoice 2
Our newest and most expressive model, with the most natural Arabic delivery. It's the one to reach for when emotion and performance matter: ads, storytelling, characters.
Its output varies from take to take: the same text can come back with different emphasis or pacing each time. Generate a few takes and keep your favorite. Use "quality": "v2".
MenaVoice 1.5
Richer, studio-style delivery with a polished finish, made for narration, e-learning and commercials. Use "quality": "pro".
MenaVoice 1v
Fast and consistent, and the default. A dependable choice for everyday scripts, product audio and anything you generate at volume. Use "quality": "flash", or leave quality out.
MenaVoice Clone
Speaks with a voice from your voice library: a voice you cloned from a recording, or one you saved from the community. Dialect steering doesn't apply, because a cloned voice keeps the accent of its sample. Use "quality": "clone" with a library voice's ID.
Speech to Text model
MenaVoice Speech to Text transcribes MP3, WAV, WebM, OGG, M4A and FLAC recordings. It's tuned for Arabic, detects the language on its own, and also transcribes English and French. See Speech to Text.
Languages
| Model | Languages |
|---|---|
| MenaVoice 2, 1.5, 1v | Modern Standard Arabic and 22 regional dialects, plus English and French |
| MenaVoice Clone | The language of your text, in the accent of the cloned voice |
| Speech to Text | Arabic, English and French, with automatic language detection |
Pricing
Every model costs the same:
- API: $25 per million characters for text to speech, and $25 per million transcript characters for speech to text.
- Web app: each character of text uses one character of your plan's monthly quota, whatever the model. Transcriptions use your plan's speech-to-text minutes.