Language experts meet broadcast workflows: Intro­ducing Arabic speech-to-text with Lisan and DeepVA

In today’s information-driven world, spoken content is one of the most valuable assets for organi­za­tions and insti­tu­tions. Parlia­mentary sessions, events, meetings and broad­casters and news outlets all generate hours of audio and video content every day. Capturing, transcribing and under­standing this data is important not only for efficiency, but also for trans­parency, account­ability and acces­si­bility.

Tradi­tionally, manual transcription is time-consuming and resource intensive. Tradi­tional open-source or hyper­scaler solutions are tied to cloud usage and token-based pricing, and often require hosting and mainte­nance by in-house staff. However, the bigger problem for regional markets is transcribing dozens of regional dialects under the general term ‘Arabic’.

Why is Arabic speech-to-text different?

In practice, Arabic is not one language, but a family of languages comprising over 32 dialects, each with its own vocab­ulary, meaning, grammar, and pronun­ci­ation. From Casablanca to Baghdad and Cairo to Riyadh, Arabic speech varies dramat­i­cally. Tradi­tional systems often fail to capture this diversity, resulting in high error rates and unreliable transcripts for local markets.

This is where Lisan’s Voice model comes in — it is an industry-leading speech-to-text system that has been specif­i­cally designed for the Arabic language and its complex­ities. When combined with DeepVA’s transcription and semantic enrichment capabil­ities, you can generate high-quality Arabic and English transcripts in real time, and all provided using broadcast crate infra­structure and standards.

USPs & facts

  • Multi­lingual audio processing:

    Handles both Arabic and English segments seamlessly.

  • Dialect coverage:

    Supports over 32 Arabic dialects — from Cairene to Casablanca and from Doha to Damascus.

  • System integration:

    Being available in DeepVA allows for straight­forward integration with any existing broadcast workflow via a single API.

  • On-premises deployment:

    Maximum confi­den­tiality is ensured for sensitive insti­tu­tions such as parlia­ments, ministries, broad­casters and big media outlets.

Intro­ducing compre­hensive dialect support

Our collab­o­ration with Lisan AI has been enhanced to deliver industry-leading speech recog­nition across 32 distinct Arabic varieties, covering virtually every major urban centre and regional variant in the Arab world. This includes both founda­tional forms, Classical Arabic and Modern Standard Arabic, alongside compre­hensive regional dialect support spanning:

Levantine varieties: From Damascus and Aleppo to Beirut and Jerusalem, our system captures the nuanced differ­ences in Levantine Arabic, including urban variants such as the dialects of Hebron and Amman.

Gulf States Coverage: Complete support for Gulf Arabic varia­tions, including Kuwaiti, Bahraini (Manama), Qatari (Doha) and the distinct Emirati dialects, as well as Saudi variants from Jeddah to Riyadh.

North African integration: We provide full coverage of Maghrebi Arabic variants, from Casablanca and Rabat in Morocco to Tunis and Sfax in Tunisia. We also offer compre­hensive coverage of Algerian variants, including the dialects of Algiers and Oran.

Mesopotamian and Eastern Coverage: Support for Iraqi variants, including Baghdadi, Maslawi and Basrawi, as well as Yemeni dialects from Sana’a and Aden.

Regional Specialists: Even less commonly supported variants such as Sudanese (Khartoum), Libyan (Tripoli, Benghazi and Sabha) and specialized urban varieties are now fully integrated.

Beyond transcription: Produc­tivity & Gover­nance

Lisan AI is a deep-tech AI company, special­izing in multi­lingual writing assis­tants, meeting management solutions, document automation, and produc­tivity tools tailored for enter­prises and govern­ments in the MENA region. Founded with the mission of bridging global AI advance­ments to local market needs, Lisan’s products address challenges in language gover­nance, compliance, commu­ni­cation efficiency, and insti­tu­tional produc­tivity in their region and beyond. Providing this knowledge via a versatile, broadcast-focused DeepVA infra­structure via a single API brings these advan­tages into real-life production workflows, even when required on-premises.

Why this matters

By integrating DeepVA’s Lisan’s voice model into our Speech Recog­nition, we are bridging the gap between global AI innovation and regional linguistic realities. The result:

  • Faster and more accurate transcription.

  • Greater inclu­sivity for Arabic-speaking commu­nities.

  • Reliable, confi­dential and scalable solutions for broad­casters according to their standards and workflows.

  • Highest Security require­ments are matched thanks to the ability of hosting it entirely locally.

  • Thanks to fixed pricing, avoid token-based pricing and make only the hardware infra­structure the limiting factor.

But this isn’t just about produc­tivity. It’s about making an impact: empow­ering journalists, media companies and insti­tu­tions to work faster, deliver more accurate content and, ultimately, strengthen democ­ratic trans­parency through technology, regardless of the language.

With DeepVA and Lisan AI, Arabic languages are no longer a barrier.
They become a bridge.

Intereset in learning more? Join us at the IBC and experience it in action.

Share

Email
LinkedIn
Facebook
Twitter
Search

Table of Contents

latest AI news

Subscribe to our newsletter

Don’t worry, we reserve our newsletter for important news, so we only send a few updates once in a while. No spam!