photo: 24kg
Kyrgyzstan has made a strong statement on the global tech stage at CES 2026, unveiling homegrown artificial intelligence solutions that rival leading international platforms.
The centerpiece of the national showcase was KaniTTS, an open-source text-to-speech model developed by Kyrgyz AI startup NineNineSix, The Caspian Post reports via Kyrgyz media.
The system generates natural, real-time speech and is reported to be up to three times faster and ten times cheaper than top commercial services from companies such as ElevenLabs, OpenAI and Google.
Licensed under Apache 2.0, KaniTTS can produce 15 seconds of high-quality speech in just one second using a consumer-grade NVIDIA RTX 5080 graphics card, eliminating the need for costly cloud infrastructure. Since its launch, the model has been downloaded more than 15,000 times on Hugging Face, reflecting strong global interest. It already supports eight languages, including Kyrgyz, English, German, Korean, Arabic, Chinese, Spanish and Japanese.
Alongside KaniTTS, the Kyrgyz team introduced Kyrgyz Whisper, an open-source speech recognition model fine-tuned from OpenAI’s Whisper. Trained on around 2,000 hours of Kyrgyz speech, the model reduced recognition errors for the Kyrgyz language from nearly 100 per cent to just 0.2 per cent, addressing a major gap in AI support for underrepresented languages.
Thanks to its open architecture, KaniTTS is already being adapted worldwide, with developers building new voice solutions for languages such as Vietnamese, Urdu, Creole and Hausa, expanding access to modern AI tools for communities often overlooked by major tech corporations.
Kyrgyzstan’s presence at CES 2026 underscored that the country is not only exporting IT services but also creating world-class technology products. Held annually in Las Vegas, CES is the world’s largest technology exhibition, bringing together thousands of companies, startups, investors and media from over 150 countries to showcase the future of innovation.
Share on social media