Zol AI Lab
← All projects
Product · since October 2026live

ZolVoice

Natural Sudanese Arabic voices from text.

A text-to-speech product for Sudanese Arabic, live since 13 October 2026 and used by thousands of creators, ad designers and podcast owners.

Content creatorsAd designersPodcast owners
Visit zolvoice.com →
Livesince 13 October 2026
88%support-agent pass rate, up from 27%
~990entries in the Sudanese lexicon

How it works

From input to finished result.

1
Your scriptSudanese Arabic or English, short or long
2
PronunciationSudanese lexicon, homograph rules and AI vowel marking
3
RoutingEach script scored for difficulty and sent to the right voice model
4
FailoverAutomatic switch when a model is busy or slow
5
Natural voiceLong scripts voiced in parallel and joined

What it does

The parts that make ZolVoice useful.

Studio

Write, voice and download, in English and Arabic (right to left), with streamed progress for long scripts.

English · Arabic (RTL) · streamed progress

Pronunciation layer

Fixes how Sudanese words are spoken before the voice model sees them; a nightly scout proposes new fixes for approval.

~990-entry lexicon · AI vowel marking

AI support agent

Reads support email, finds the answer, checks the account read-only, replies in the customer's dialect, and waits for a person to approve.

Hybrid search · guardrails · human approval

Admin console

Real cloud cost, latency, errors, quotas and an audit log in one place.

Real cloud spend · latency · audit log

Models and tools

What ZolVoice runs on.

Names and logos belong to their owners and identify the models and tools we use; no endorsement is implied.

Built so far

Engineering that is already in place.

  • Support agent: pass rate from 27% to 88% with zero forbidden claims
  • A blind audio judge scores speech quality without seeing the target text
  • Spend guard pauses generation if usage spikes
  • Layered abuse controls on free credits

The research behind it

Lab work that shaped the product.

  • Built 6,000+ clean training clips from 22 hours of speech
  • Multi-model speech-recognition voting halved transcript error from 31.5% to 15.3%
  • Fine-tuned XTTS-v2 and F5-TTS voices with held-out evaluation

Lab notes

Progress on ZolVoice.

ZolVoice is live

ZolVoice, our Sudanese Arabic text-to-speech product, is live at zolvoice.com. It routes each script to the right voice model, fixes Sudanese pronunciation before speaking, and answers support emails with an AI agent that a person approves.

product · speech

Designed and built at Zol AI Lab by Moe Eldouma, founder. See his role, skills and other work on eldouma.dev.