The AI startup NineNineSix has released KaniTTS, a groundbreaking text-to-speech (TTS) model that supports Arabic alongside other major world languages - and it’s completely free and open-source.
While global giants like OpenAI, Google, and Microsoft have dominated voice AI, KaniTTS brings high-quality, expressive Arabic speech synthesis to everyone-developers, startups, educators, and creators - at zero cost.
KaniTTS can generate 15 seconds of human-like, emotionally rich speech in just 1 second, running smoothly on a regular NVIDIA RTX 5080 GPU — no cloud servers or expensive licenses required.
It currently supports Arabic, English, German, Korean, Chinese, and Spanish, with Kyrgyz and Japanese coming soon. The model has already been downloaded over 15,000 times on Hugging Face, one of the world’s largest AI platforms.
“Our goal was to democratize voice AI — to make it open, fast, and expressive in every language, including Arabic,” said the NineNineSix team. “We believe everyone should have access to world-class AI technology, not just big corporations.”