HUMAIN Launches Arabic-First Voice AI Platform

2 Min Read

Saudi AI company HUMAIN has launched HUMAIN Voice, a platform that brings speech recognition, text-to-speech and voice agents together, with Arabic dialects and Arabic-English code-switching built into its design.

The platform supports Saudi dialects including Najdi and Hijazi, as well as Khaleeji and other Arabic dialects, English, and switching naturally between Arabic and English. HUMAIN says the service is designed for customer service, enterprise operations and other uses that rely on spoken conversations.

Its voice agents are full-duplex speech-to-speech systems that can listen and respond in a caller’s dialect, which HUMAIN says they do with low latency. The company says its HUMAIN Voice API is available to developers and is already in use.

HUMAIN also introduced HUMAIN-Voice-Lite-80M, an 80-million-parameter automatic speech recognition model covering Arabic, English and Arabic-English code-switching. The model is designed to run on CPUs as well as GPUs, supporting deployments where cost, infrastructure or on-device and edge use are considerations. HUMAIN says it is making the model’s weights available under the Apache 2.0 licence so developers, researchers and enterprises can adapt the technology.

The company says it will publish benchmark results from the Open Universal Arabic ASR Leaderboard alongside a blog demonstrating the model across Arabic, English and code-switching speech examples. HUMAIN describes Voice as an extension of its portfolio into the interfaces people use to interact with AI.

Source: MiddleEastAI

Share This Article