AI

NCSpeech: the startup teaching AI to understand speech

A microphone and a sound wave on a screen representing speech recording for voice model training

Изображение: NCSpeech

NCSpeech is a Kazakhstani startup that collects speech recordings and builds models for speech recognition and voice synthesis. The company says it had collected 550 hours of speech from 2,704 speakers by November 21, 2026 (ncspeech.org).

The company turns those recordings into materials for AI developers. It also offers ready-to-use tools for working with Kazakh and Russian speech through NCSpeech Studio.

What NCSpeech does and how it works

A voice model needs examples of how people speak in real life. They might switch languages mid-sentence or record their voices with noise in the background. NCSpeech collects and checks such recordings through a mobile network of participants.

The company licenses datasets: collections of recordings and transcripts used to train and test models. Its catalog includes banking requests and medical phrases. It also has separate collections of names and addresses, spontaneous conversations, whispers and background noise.

NCSpeech also uses the recordings to train its own speech models. In October, Astana Hub reported that the company had introduced models for mixed Kazakh-Russian speech developed jointly with inDrive. They are available in NCSpeech Studio (astanahub.com).

How NCSpeech collects recordings and how Studio works

The founders say the company has its own multi-agent platform. It assigns tasks to participants and checks the results. The system sends poor-quality work back to the person who submitted it for correction (digitalbusiness.kz).

On the Studio website, users can upload audio or video and get a transcript with timestamps and speaker labels. They can download it as text, subtitles or a JSON file, a format programs use to exchange information (studio.ncspeech.ai).

The service also works in reverse, turning Kazakh or Russian text into audio in a selected voice. Users can create their own voice from a sample recording. Studio is available in a browser and through an API, an interface for adding its features to another service.

The company also describes other voice products. Voice Control Layer is designed to control apps through speech, while Voice AI Assistant is intended for customer support and workflow automation (ncspeech.ai).

Who uses the products and how NCSpeech makes money

Companies buy the datasets when they need material to train their own models. In May 2026, the founders said they had several clients in Kazakhstan and Southeast Asia. They included banks and other financial institutions, as well as AI labs within holding companies.

The team said orders started at tens of thousands of dollars. The price includes payments to the people collecting recordings and a share for the partner app. The company said it was operationally profitable but did not disclose its revenue.

Studio has its own pricing model: paid subscriptions renew monthly, and usage is measured in credits. Standard recognition of one second of an audio channel costs one credit, while synthesizing one second of speech costs 12 credits. Creating a custom voice costs 100 credits, according to Studio’s terms.

Studio can be used to transcribe meetings and interviews. It can also provide voiceovers for videos and educational materials. Developers can add speech features to their products through the API.

How many recordings NCSpeech has collected and what support it has received

According to NCSpeech, its catalog contained 15 datasets with recordings from four countries as of November 21, 2026. They covered five languages: Spanish and Kazakh, as well as Malay, Portuguese and Russian. The company counted 128,853 utterances with text transcripts.

Those figures show the volume of recordings collected. They do not reflect the number of paying customers. The company also says it can organize data collection in 45 markets.

In May, the founders said they had raised $110,000 from Antler. The fund lists NCSpeech in its portfolio as a project from Malaysia but does not disclose the investment amount (antler.co). Information about the company is also available in the NCSpeech profile.

Akorda confirmed that NCSpeech won the Future Builders category at the Alem.ai Battle competition in April 2026. The founders put the prize at 10 million tenge, though Akorda’s announcement did not specify the amount (akorda.kz).

In the fall, the startup also won the Market Shaper category at Astana Hub Battle. The organizer said each of the four winners received $25,000. This was a separate award from a different competition.

Who founded NCSpeech

The project’s founders are Tseren Andzhukaev, Nuraly Zhanbyrbaev, Dmitry Sandzhiev and Artem Ivanov. The company says the project was created in March 2025. The team initially worked on voice products, then focused on collecting and preparing recordings.

Andzhukaev leads the company, while Ivanov is responsible for technology and machine learning infrastructure. Sandzhiev develops partnerships in Southeast and Central Asia, as well as the Middle East and North Africa. Zhanbyrbaev runs operations in Central Asia.

NCSpeech says its core team has worked together since 2018. The founders funded the project themselves for its first six months. The company lists entities in Kazakhstan and Malaysia on its website.

What’s next for NCSpeech

In May, the team planned to raise about $3 million and expand its development team. The startup also intended to hire for engineering and commercial roles. The founders wanted to develop NCSpeech as a supplier of model training materials for emerging markets.

The Kazakh and Russian speech models are already available in Studio. Astana Hub reports that NCSpeech continues to work with inDrive, and app users can take part in collecting speech recordings.

Startup Radar: NCSpeech

Понравился материал? Поделитесь с другими:

Picture of Раимбек Искендиров
Раимбек Искендиров
Picture of Раимбек Искендиров
Раимбек Искендиров

Похожие материалы

A loan agreement with the annual effective rate line highlighted and a calculator beside it.

Reviews

Insurance and collateral valuation costs will count toward APR on loans in Kazakhstan from November 30

From November 30, Kazakhstan changes how APR is calculated and disclosed. See which costs ...
A loan agreement and a calculator lying next to an income statement.

Reviews

National Bank puts off its 8-years-of-income debt cap

The National Bank has delayed its cap on debt at eight years of income. ...
Close-up of cloud server racks, the infrastructure startups received AI credits to use.

Reviews

Alibaba Cloud awards $40,000 in AI credits to startups in Kazakhstan and Uzbekistan

Alibaba Cloud has awarded $40,000 in AI credits to 20 startups in Kazakhstan and ...

Читайте нас в Telegram и Instagram

Подписывайтесь на наши каналы, чтобы первыми получать самые свежие новости казахстанской IT-индустрии, анонсы грантов, стажировок и эксклюзивные интервью.