Signs → text
The camera recognizes signs and shows the text right away. A language model turns the words into a fluent sentence in Uzbek.
- Real time
- In the browser
- Russian Sign Language
- Video is not stored
Sign language → text · LIVE
isoraAI recognizes signs from a camera, builds a sentence in Uzbek and shows the staff reply in sign language.

Signs
Sentence in Uzbek
Salom, menga pasport kerak.
deaf people in Uzbekistan, 21,000 of them registered with the Society of the Deaf
sign language interpreter with a university degree at the Society of the Deaf of Uzbekistan
the year from which public bodies must provide sign language interpretation to citizens
A deaf person cannot handle a bank visit, an airport check-in or a doctor's appointment alone.
Sources: Society of the Deaf of Uzbekistan; yuz.uz; Decree of the President of the Republic of Uzbekistan (2024).
The camera recognizes signs and shows the text right away. A language model turns the words into a fluent sentence in Uzbek.
The staff member picks a phrase, and a 3D avatar signs the reply.
| Live interpreter | Video interpreter | isoraAI | |
|---|---|---|---|
| Availability | Working hours, by appointment | Working hours, by appointment | 24/7, no waiting |
| Scale | One interpreter per service point | Also limited by people | Any number of points |
| Text translation | No | No | Yes, both ways |
Public service centers, ministries, agencies
Branches and mini-banks
Ticket desks, check-in, gates
Reception and appointments
Estimate: about 57 branches per 100,000 adults (IMF, 2024).
Next, the whole region: Kazakhstan and the CIS. Russian Sign Language is shared across most of the CIS.
Measured on the open Slovo dataset, with people the model had never seen. This is Russian Sign Language; accuracy for Uzbek Sign Language is still being tested.
A team of four and five experts from the deaf community.
isoraAI is a digital sign language translator. A tablet camera sees the sign, the server recognizes it and builds a sentence in Uzbek. The staff reply is shown by a 3D avatar or a sign video.
The current models are trained on the open Slovo dataset, which is Russian Sign Language. It is shared across most of the CIS. Work on adapting them to Uzbek Sign Language is under way.
No. Frames are processed only during the conversation and are not stored.
A tablet or computer with a camera and an internet connection. isoraAI runs in the browser, so there is nothing to install. The heavy computing happens on the server.
For now the demo is open only to users who were given a login and password. After signing in, the session stays active for a long time.
Sign in with your login and password: camera, live translation and the 3D avatar on one page.
Open the demo