Talksign AI Models Bridge Communication Gap for Deaf Users

Talksign AI Models Bridge Communication Gap for Deaf Users

Quick Reads

• Talksign launched Palm 1.0 and Echo 1.0 on May 20, both enabling real-time ASL translation.
• Palm 1.0 translates sign language into text or speech with 84.2% semantic accuracy.
• Echo 1.0 converts written or spoken language into photorealistic ASL video at 30 frames per second.
• The models were built on a database of over 71,000 to 94,000 ASL samples reviewed with Deaf advisors.
• A full rollout on desktop and Meta Ray-Ban smart glasses is scheduled for August 20, 2026.

Talksign sign language AI has taken a major step forward through the Nigeria- and UK-based startup. Specifically, the company launched two models on May 20, 2026. Together, the models aim to improve communication between Deaf and hearing people at scale.

Talksign introduced Palm 1.0 as its sign-to-text model. In particular, the model translates American Sign Language into written or spoken language in real time. It achieves 84.2% semantic accuracy. Additionally, it records 79.6% word-level accuracy. The company trained the model on more than 71,000 ASL samples. Furthermore, it used a transformer-based architecture called SAGE. SAGE stands for Spatial Attention Graph Encoder. Notably, the system tracks about 133 anatomical landmarks across hands, heads, and shoulders. As a result, the design interprets continuous signing as contextual conversation. Instead, it does not simply recognise isolated gestures.

Meanwhile, Talksign developed Echo 1.0 for the opposite direction of communication. The model converts written or spoken English into photorealistic ASL video. Moreover, the avatars render at 30 frames per second. The system also maintains latency of about 29 milliseconds. In addition, users can generate personalised avatars from one reference photo. According to the company, the feature creates a more natural signing experience. Consequently, it aims to avoid robotic movement.

Meanwhile, Edidiong Ekong described Palm 1.0 as the first deployable large-scale model for Deaf users. He also outlined plans for deployment on phones and smart glasses. Furthermore, he mentioned future use in classrooms and hospitals.

The Talksign AI models build on the earlier Talksign-1 foundation model. Previously, the company released that version in February 2026. It provided basic translation of 250 isolated ASL signs. However, it could not interpret continuous sentences or fingerspelling. Therefore, Palm 1.0 and Echo 1.0 specifically target those limitations.

According to World Health Organization, over 430 million people worldwide have disabling hearing loss. However, most digital platforms still assume users can hear and speak. As a result, Deaf users remain excluded from everyday digital participation. Consequently, Talksign directly aims to close that communication gap.

Currently, both models support only English. However, the company plans to add Spanish, French, and Arabic in future versions. Echo 1.0 also partially supports multi-word ASL phrases. Nevertheless, the company still needs more fine-tuning for medicine, law, and engineering vocabulary. Meanwhile, Talksign identifies privacy as a major priority. The user’s device performs landmark extraction locally. Therefore, the system transmits processed data points instead of raw video.

Finally, Talksign has scheduled a full desktop and Ray-Ban Meta Smart Glasses rollout for August 20, 2026.