Kazakhstani startup NCSpeech, a member of Astana Hub, has introduced voice models developed in collaboration with inDrive. According to the team, its model makes half as many errors when recognising Kazakh speech as solutions from the world’s leading technology companies. The technology also recognises mixed Kazakh-Russian speech, reflecting everyday communication patterns in Kazakhstan.
The models were developed using hundreds of hours of natural speech recorded by inDrive users in exchange for compensation. The dataset includes recordings in Kazakh and Russian, as well as conversations involving switches between the two languages. This allowed the team to account for natural speech patterns, such as starting a sentence in one language and finishing it in another.
“In Kazakhstan, people switch freely between Kazakh and Russian, sometimes within a single sentence. We wanted to create technology that understands this natural way of speaking. Our collaboration with inDrive helped us collect hundreds of thousands of hours of recordings and improve the quality of our models. It is important to us that people can use voice services in the language they are comfortable with, without having to adapt their speech to the system’s capabilities,” said Tseren Andzhukaev, Director of NCSpeech.
Recognising mixed-language speech remains a challenge for many voice systems. NCSpeech’s technology expands the potential applications of voice AI in Kazakhstan, including call centres, subtitling services, applications for blind users and other products where understanding users accurately is essential.
The voice models are already available in NCSpeech Studio. The platform enables users to transcribe recordings of meetings, interviews or lectures, convert text into speech in Kazakh or Russian, create a custom voice from a short recording and integrate the technology into their own products.
The NCSpeech team has already earned recognition within the technology ecosystem: in 2026, the startup won alem.ai battle and received the Market Shaper award in the Astana Hub Battle competition at the international AI & Digital Bridge 2026 forum.
NCSpeech’s experience demonstrates how Astana Hub members are creating competitive AI products that address local market needs. The startup’s collaboration with inDrive is ongoing, and any user of the app can contribute to the collection of speech data.
Explore the platform’s capabilities at NCSpeech Studio.
Kazakhstani startup NCSpeech, a member of Astana Hub, has introduced voice models developed in collaboration with inDrive. According to the team, its model makes half as many errors when recognising Kazakh speech as solutions from the world’s leading technology companies. The technology also recognises mixed Kazakh-Russian speech, reflecting everyday communication patterns in Kazakhstan.
The models were developed using hundreds of hours of natural speech recorded by inDrive users in exchange for compensation. The dataset includes recordings in Kazakh and Russian, as well as conversations involving switches between the two languages. This allowed the team to account for natural speech patterns, such as starting a sentence in one language and finishing it in another.
“In Kazakhstan, people switch freely between Kazakh and Russian, sometimes within a single sentence. We wanted to create technology that understands this natural way of speaking. Our collaboration with inDrive helped us collect hundreds of thousands of hours of recordings and improve the quality of our models. It is important to us that people can use voice services in the language they are comfortable with, without having to adapt their speech to the system’s capabilities,” said Tseren Andzhukaev, Director of NCSpeech.
Recognising mixed-language speech remains a challenge for many voice systems. NCSpeech’s technology expands the potential applications of voice AI in Kazakhstan, including call centres, subtitling services, applications for blind users and other products where understanding users accurately is essential.
The voice models are already available in NCSpeech Studio. The platform enables users to transcribe recordings of meetings, interviews or lectures, convert text into speech in Kazakh or Russian, create a custom voice from a short recording and integrate the technology into their own products.
The NCSpeech team has already earned recognition within the technology ecosystem: in 2026, the startup won alem.ai battle and received the Market Shaper award in the Astana Hub Battle competition at the international AI & Digital Bridge 2026 forum.
NCSpeech’s experience demonstrates how Astana Hub members are creating competitive AI products that address local market needs. The startup’s collaboration with inDrive is ongoing, and any user of the app can contribute to the collection of speech data.
Explore the platform’s capabilities at NCSpeech Studio.
Share
