Voice & speech

NVIDIA open-weight batch speech model transcribing 40 languages, including Arabic (rated "Supported"; 6.35% word error on the FLEURS Arabic test). Handles finished recordings up to one hour, speech translation to and from English, and speaker labels. OpenMDW licence. Not live streaming.

Arabic speech recognition released alongside Jais 3.

1.6B speech-recognition model transcribing Standard and Emirati Arabic, English, French, Spanish and Portuguese, with word-level timestamps. Downloadable weights and licence terms are not confirmed.

Darija speech recognition (Voxtral-based, handles Arabic–French–English mixing) and an Arabic dialect identifier.

AI sales rep that phones customers in Saudi Arabic about abandoned carts and offers.

Conversational voice AI tuned for Saudi, Maghrebi, Egyptian and Levantine dialects.

Free Egyptian and MSA transcription, translation to 50 languages and Arabic text-to-speech.

Arabic voice AI platform, deployed in Saudi Arabia with OmniOps on sovereign infrastructure.

2B speech-to-text for MSA, Gulf, Egyptian, Levantine and Maghrebi, with code-switching.

Saudi voice AI platform offering Arabic speech-to-text, text-to-speech and voice agents, with API access.

OpenMOSS / MOSI.AI 0.1B speech model with Arabic (ar), CPU streaming and voice cloning. Public downloadable weights; no hosted Hugging Face Inference Provider. Apache-2.0 in GitHub and model metadata, but the model card retains a redistribution warning. Official language lists differ; Arabic is confirmed. Arabic audio quality not tested by Souq.

150M Arabic–English text-to-speech with voice cloning; commercial voices at app.silma.ai.

Bilingual speech-to-text for conversations that switch between Arabic and English, including within a sentence.

Voice automation built on its own dialect-based Arabic speech AI.

OpenMOSS long-form dialogue speech model with Arabic (ar), voice cloning and 1–5 speakers. Public downloadable MOSS-TTSD-v1.0 weights; no hosted Hugging Face Inference Provider. Apache-2.0 in GitHub and model metadata. Official language counts differ; Arabic is confirmed. Arabic audio quality not tested by Souq.

OpenMOSS multilingual text-to-speech and voice-cloning model family with Arabic (ar) support. MOSS-TTS 1.0 supports 20 languages; v1.5 expands to 31. Released under Apache-2.0. Arabic audio quality has not been independently tested by Souq.

OpenMOSS streaming speech model for context-aware, multi-turn voice agents, with Arabic (ar) and voice cloning. Public downloadable weights; no hosted Hugging Face Inference Provider. Apache-2.0 in GitHub and model metadata. Official language counts differ; Arabic is confirmed. Arabic audio quality not tested by Souq.

Canopy Labs’ Saudi Arabic text-to-speech model is available through the GroqCloud API.

Arabic speech-to-text model supporting Modern Standard Arabic and Gulf, Levantine, Egyptian, Maghrebi and Iraqi varieties, available through a cloud API or self-hosted deployment.

An Arabic-first voice platform offers text-to-speech, speech recognition and voice cloning.

F5-based Arabic speech synthesis.

400M Arabic-specialised text-to-speech model.

Contact-centre software with an Arabic-native AI agent and dialect call transcription.

24/7 Arabic voice agents for customer support, bookings and sales.

Unified Arabic text-and-speech transformer covering 17 dialects.

Distilled Arabic speech-recognition models provide checkpoints for audio transcription.

AI speech generation and localization tools support Arabic narration and multilingual voice production.

Agentic voice AI for customer support in MENA contact centres.

Voice AI with Arabic text-to-speech and voice agents; spoke at Web Summit Qatar 2026.

Arabic speech intelligence across 25+ dialects, plus the Ziila digital human.

Multilingual voice AI (including Darija) for customer service and call monitoring.

Avatars that translate Arabic text and speech into sign language.

Multi-dialect text-to-speech (Egyptian, Saudi, Moroccan, Iraqi) with tashkeel support.

A one-billion parameter speech synthesis model built on the Llama architecture to generate natural-sounding voice.

UAE data and AI company providing enterprise and government AI solutions, including Arabic voice agents and on-premise deployment.

Advanced systems that identify and translate spoken language into text for hands-free device interaction.

A state-of-the-art multilingual foundation model for advanced speech generation and processing.

Instantaneous linguistic processing tools that convert speech or text across different languages during live interactions.