Sudanese text to speech
Sudanese Arabic text to speech, built for how Sudan actually talks.
Sautna (صوتنا— “our voice”) turns written Arabic into speech that sounds like Khartoum, Omdurman or Wad Madani. Not a Cairo newsreader. Not Modern Standard Arabic with an accent painted on. It runs as a browser playground and behind a one-endpoint API, and the same engine is answering live phone calls today.
The problem
Why MSA and Egyptian TTS fail on Sudanese text
Every general Arabic voice can pronounce Sudanese words. That is not the same as speaking Sudanese. Four things break, and they break in the same order every time.
The consonants are wrong
Across most of Sudan, ق is a hard [g] and ج leans toward [j]. An MSA-trained model says [q] and [dʒ]. One letter, and the listener has already placed the speaker somewhere else.
The vocabulary is unfamiliar
هسع (now), شنو (what), داير (want), زول (person). A model that has never seen these in training guesses at stress and pausing, and the guess is usually wrong.
The rhythm is somebody else's
Sudanese Arabic is unhurried — longer vowels, softer landings at the end of a phrase. Egyptian TTS is quick and clipped. The words can be correct while the music is entirely foreign, which is the version people find hardest to listen to.
The training labels lied
Most Arabic speech corpora are transcribed into Modern Standard Arabic even when the speaker did not say it that way. Train on that and you get a model that reads MSA no matter what you feed it. We labelled our corpus as-spoken — that single decision is most of the difference you hear.
The voices
Zol and Zola
Both live voices are inventions. They are blends built from our corpus so that no identifiable person's voice is being sold — consent-clean by construction, which matters when your audio ends up in an ad, an IVR tree or a national broadcast.
voice id: zol
Warm and conversational — the flagship. Narration, assistants, phone lines and anything that needs to sound like a person you already know.
voice id: zola
Bright and friendly — the female voice. Announcements, onboarding flows, learning tools and customer-facing menus.
API & playground
Try it in the browser, ship it in five minutes
The demo box on the home page needs no account — type Sudanese Arabic and listen. Sign up and the studio playground lets you audition scripts, switch voices and styles, and download the takes you like. When you are ready to automate it, the same synthesis sits behind one POST request.
You get 24 kHz WAV by default, with MP3, Opus, raw PCM and 8 kHz μ-law for telephony. Set stream: true and audio starts arriving on the first finished sentence rather than the last.
curl -X POST https://sautna.com/api/v1/tts \
-H "Authorization: Bearer sautna_sk_..." \
-H "Content-Type: application/json" \
-d '{"voice": "zol", "text": "الليلة الجو سمح، والقهوة جاهزة."}' \
--output speech.wavHow it compares
What the other engines do and don't cover
To be straight about it: ElevenLabs, Play.ht, Google, Azure and Amazon Polly all make genuinely good Arabic voices, and for Modern Standard Arabic they are excellent — better tooling, bigger catalogues, more languages than we will ever have. Some now offer Egyptian and Gulf flavours too. What none of them ship, at the time of writing, is a Sudanese dialect voice. lahajati.ai is the closest thing to a dialect-first Arabic TTS and is worth your time if you need Levantine or Khaliji; Sudanese is not in its lineup either.
So the comparison is not “ours is better than theirs.” If your script is MSA, use whichever of them you already have a contract with. The comparison is that 50-odd million Sudanese Arabic speakers — in Sudan and across the diaspora in the UAE, Saudi Arabia, Egypt, the UK and the US — have never had a synthetic voice that sounds like them, and building one required a Sudanese speech corpus that did not exist until we collected it. That corpus, labelled as-spoken rather than normalised into MSA, is the whole moat.
Where it goes
IVR & call centers
Telephony-ready 8 kHz μ-law output, streaming first audio, and voices already running on live SIP lines. For banks, telecoms and Gulf contact centres with Sudanese customers. →
Creators & dubbing
Sudanese voice over for YouTube, TikTok and podcasts, plus dubbing and shareable voice notes — with a commercial licence and no booth. →
Questions
What is Sudanese Arabic text to speech?
It is a TTS engine trained on Sudanese speech rather than Modern Standard Arabic. You send Arabic text and get back audio in the Sudanese dialect — the hard [g] for ق, Sudanese vocabulary like هسع and شنو, and the slower Khartoum rhythm. Sautna is currently the only production TTS product that offers it.
Why can't I just use a normal Arabic TTS for Sudanese text?
You can, and it will read the words — but it reads them as a Cairo or MSA newsreader would. The letter ق comes out as [q] instead of the Sudanese [g], dialect words get MSA stress and pauses, and the intonation is clipped where Sudanese Arabic is long and falling. Sudanese listeners hear a foreigner reading a Sudanese script.
Which Sudanese voices are available?
Two live voices: zol (زول), a warm male voice, and zola (زولة), a bright female voice. Both are invented voices — blends built so they do not clone any identifiable person. A studio-recorded hero voice is in training for Scale customers.
Do other TTS providers have Sudanese Arabic voices?
As of now, no. ElevenLabs, Play.ht, Google, Azure and Amazon Polly all ship strong Arabic, but it is Modern Standard Arabic, with Egyptian and Gulf coverage appearing in some catalogues. lahajati.ai covers several Arabic dialects and does it well; Sudanese is not one of them. That gap is the reason Sautna exists.
Can I use the audio commercially?
Yes on any paid plan — Starter at $9/month and up include a commercial licence for the audio you generate. The free tier's 5,000 characters a month are for personal use and evaluation.
What audio formats and sample rates do I get?
24 kHz mono 16-bit WAV by default, plus MP3, Opus, headerless PCM, and 8 kHz μ-law for telephony. Streaming returns raw 24 kHz PCM sentence by sentence so playback can start before the full script finishes rendering.
Hear your own words in Sudanese
5,000 characters a month, free, no card. Paid plans start at $9/month with full API access and a commercial licence.
Start free