01 · Recorded
A person, in a room, in their dialect
Not a dataset, but voices of people from Sharjah, Jeddah, Kuwait and Cairo, professionally recorded and speaking as they do at home.
Voice Library
Every voice has a hometown.
Twenty voices available now, across eleven ways of speaking: eight Arabic dialects, plus Fusha and American and British English. Each voice was recorded by a speaker from its own community. Choose one, listen and use it.
20voices today
11ways of speaking
Abdulla.
Emirati
A professional voice for a range of use cases.
The library
Twenty voices, by where they come from.
One list. Click a row to hear it.
Emirati4 voices
Saudi1 voice
Egyptian2 voices
Kuwaiti4 voices
Fusha2 voices
American4 voices
British3 voices
Qataricoming soon
Being recorded now. The first voices land here when they are ready.
Omanicoming soon
Being recorded now. The first voices land here when they are ready.
Lebanesecoming soon
Being recorded now. The first voices land here when they are ready.
Syriancoming soon
Being recorded now. The first voices land here when they are ready.
A game
Guess the dialect.
Five voices and five guesses. Each voice says the same sentence in its own dialect; your task is to identify where it comes from. Most people from the region get four out of five. Systems that support Arabic get none, because they treat them all as one language.
Guess the dialect · round 1 of 5
Press play, then say where this voice is from.
Listen, then pick a dialect
Where they come from
Voices from the people who own them
Real people. Their voice stays theirs.
We do not use recordings scraped from the internet. Every voice in the library was recorded by someone from its own community under a signed agreement, and its owner retains the rights. Whenever you use the voice here, its owner is paid. Want to add your voice to the library? Use Voice Cloning.
02 · Agreed
Signed, scoped, theirs
A written agreement says what the voice may be used for. The person can see where it is used and can withdraw it.
03 · Paid
Used here, paid there
Whenever a voice is used, its owner earns from it. Ongoing income from their voice, not a one-time payment. Coming Soon
Developers
List them. Pick one. Speak.
The library is an endpoint. List the voices, take a voice id, hand it to Text to Speech or to a Sada agent.
# python import requests API = "https://api.echonova.ai" AUTH = dict(Authorization="Bearer YOUR_API_KEY") voices = requests.get(API + "/v1/text-to-speech/voices", headers=AUTH) ahmed = next(v for v in voices.json()["data"]["items"] if v["name"] == "ae_ahmed") res = requests.post( API + "/v1/text-to-speech/sync", headers=AUTH, json=dict(text="هلا والله", voiceId=ahmed["id"]), )
Security & sovereignty
Built in the UAE. Your audio stays in the UAE.
Your channels
Phone calls
Customer service agent interface
Apps & websites
Recordings & files
Audio comes in, text and spoken responses go out. Your data never leaves the region.
Speech to TextConvert audio to text in real time or in batches, across eight dialects.
Text to SpeechNatural Arabic voices in the caller’s own dialect.
Sound Cleaning
Noise and echo are removed before the model listens to the audio.
Your systems
Contact centre platform
Customer relationship management and accounts
Your own code
Analytics & QA
From calls, apps or recordings.
Data residency and retention policyNo cross-border transfer
Common questions
How do I choose an Arabic voice?
Listen to the available samples, compare dialects and choose a voice that matches the audience and content.
Where can I use a selected voice?
Use the Text to Speech demo or follow the API documentation to generate speech with an available voice.
Sovereign Arabic voice AI
Developed in the UAE to speak the language of the region.
Try Speech to Text, Text to Speech and Sound Cleaning, or talk to us about Sada for your contact centre.