Nigerian-based AI startup model translates American Sign Language in under 100 milliseconds
Talksign, a Nigeria and UK-based AI company, launched Talksign-1 on Monday, a model it says translates American Sign Language into speech and text in under 100 milliseconds.
Reporting from the company says the release targets a structural bottleneck in digital infrastructure: most mainstream interfaces assume spoken audio input and output.
World Health Organisation figures put the addressable need at scale, with over 430 million people worldwide who are deaf and 70 million people who use sign language as their primary communication method, yet core tools like video conferencing and service kiosks rarely treat sign language as a first-class input.
Interface control shifts toward foundation layers
Talksign framed Talksign-1 as its first foundation model for sign language understanding. The model recognizes 250 ASL signs and works in two directions: it converts sign language captured by a webcam into speech and text, and it converts spoken or typed words into sign-language video sequences. That bidirectional design positions the system as an interface layer that can sit between users and existing apps, rather than as a standalone accessibility add-on.
Performance claims define adoption boundaries
Talksign said Talksign-1 analyzes about one second of signing before producing predictions, which the company described as a speed and accuracy tradeoff. In testing, it reported 84.7% accuracy on single signs. The company also set clear constraints. The system does not support continuous sentence-level interpretation or fingerspelling, so it currently fits use cases that tolerate isolated-sign translation rather than full conversational flow.
Data pipelines and privacy choices shape trust
Talksign said it trained the model on the WLASL2000 dataset, a large-scale ASL collection. The company said landmark extraction occurs on a user’s device in the browser and sends processed data points, not raw video, to its servers for analysis, per its privacy policy. Talksign said it partnered with deaf educators, native ASL signers, and accessibility advocates during development, and it cautioned against using the tool as a sole authority in medical, legal, or safety-critical settings without human oversight.
Talksign was founded in November 2025 by Edidiong Ekong and AI engineer Kazi Mahathir Rahman. The company said it plans to add British Sign Language, French Sign Language, and other languages while expanding beyond 250 signs.
