A 15-year-old student in Kazakhstan is developing Tusin, an AI platform that turns speech and text into sign-language videos, including facial expressions and hand movements. Its early testing with deaf and hard-of-hearing people puts a practical accessibility challenge ahead of the usual AI demo: whether the result communicates clearly to the people it is meant to serve.
Watch Desk analysis
What happened
Sanzhar Yerezhepov developed the project after seeing students at a nearby school communicate in sign language. He is now working with the Kazakhstan Society of the Deaf and sign-language interpreters to test an early version, he told the Times of Central Asia.
Yerezhepov says the team checks how accurately signs are reproduced and has demonstrated the technology to deaf and hard-of-hearing people. The project currently works with Kazakh and Russian speech, translating into Russian Sign Language. He says it can recognise spoken English too, but would also translate that into Russian Sign Language.
Why it matters
Subtitles are not an adequate substitute for everyone. Yerezhepov notes that sign languages have their own grammar, and that some people whose primary language is sign language may find written text difficult to understand. The project’s focus on facial expressions and context reflects the fact that signing conveys more than a sequence of hand movements.
That makes testing with signers and interpreters central, not decorative. A video can look impressively lifelike and still get the meaning wrong. Tusin is an early project, not a demonstrated solution at scale, but its stated approach puts the people who will use it into the testing process.
Our read
The most promising part is not the photorealism. It is the effort to check the output with deaf people and interpreters, where accuracy matters more than a convincing avatar. The Times of Central Asia reports that five organisations have expressed interest; that is a sign of potential demand, not proof the platform is ready for deployment.
What to watch
- Whether testing leads to published evidence about accuracy and comprehension.
- Which sign languages Tusin can support beyond Russian Sign Language.
- Whether the early project becomes a service that organisations can use in practice.
Discussion spark: Should sign-language AI be judged mainly by how natural its videos look, or by independently tested comprehension with the people who rely on it?
Sources and evidence
- Source update (16 October 2026, 21:31 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.