Published
10/06/2025, 17:05Bishkek, 6 October 2025 — Kyrgyz startup NineNineSix has unveiled KaniTTS, a text-to-speech (TTS) system that can compete in terms of quality and speed with solutions from giants such as OpenAI, Google, Microsoft, and ElevenLabs.
KaniTTS creates realistic, emotionally expressive speech in real time. The model has already been downloaded more than 15,000 times on the Hugging Face platform, where developers from around the world find and launch AI models in just a couple of clicks.
The main difference between KaniTTS and its competitors is its speed and naturalness. The model generates 15 seconds of speech in 1 second and understands intonation, pauses, and emotions, making it suitable for voice assistants, chatbots, games, films, and educational applications. And all this works on a regular computer with NVIDIA RTX 5080, without server accelerators.
KaniTTS currently speaks English, German, Korean, Arabic, Chinese, and Spanish. The developers will soon add Kyrgyz and Japanese. The model is completely open and available under the Apache 2.0 licence so that researchers and developers from around the world can use it for free.
‘We wanted to create a tool that democratises voice AI. Now, not only large corporations but also small teams will be able to work with technologies that used to cost millions,’ NineNineSix notes.
The launch of KaniTTS was a landmark event for Kyrgyzstan's IT ecosystem. It is the first world-class model of its kind created in the country, which has already attracted the attention of the international community.
Where it is used:
About NineNineSix:
A start-up from Kyrgyzstan that develops AI systems, generative models, and human-machine interaction technologies. The team specialises in high-performance open-source solutions in the field of speech, language, and multimodal systems.
