On October 13, 2026, Astana Hub reported that Kazakhstani startup NCSpeech introduced voice models developed in collaboration with inDrive (astanahub.com).
Capabilities of the Updated NCSpeech Models
The developers state that the system recognizes Kazakh, Russian, and mixed speech with natural transitions between languages in a single conversation. Available in NCSpeech Studio, the models support transcription of meetings, interviews, and lectures, text-to-speech in Kazakh and Russian, voice cloning from audio samples, and product integration (astanahub.com). The startup’s core business model is profiled in NCSpeech: как казахстанский стартап зарабатывает на сборе данных для чужих нейросетей.
Bilingual Speech Recognition in Practice
Processing mixed Kazakh-Russian speech with code-switching is widely needed in everyday business communications, meeting transcripts, and customer support services across Kazakhstan. A broader analysis of the local voice technology sector is available in ИИ-трансформация 2026: Аналитика рынка голосовых агентов в Казахстане и перспективы мультимодальных систем.
Testing Accuracy on Real Audio
Because the release presents the developers’ own specifications without third-party benchmarks (astanahub.com), organizations should test recognition and synthesis accuracy on their own recordings while ensuring a lawful basis for processing voice and personal data.
