AI research firm Talksign has announced the launch of two advanced models, Palm 1.0 and Echo 1.0, designed to enable real-time translation of American Sign Language. These tools aim to bridge the communication gap for millions of Deaf and hard-of-hearing individuals by providing a production-ready solution. The Nigeria- and UK-based company's new platform marks a significant step in deploying accessible and sophisticated sign language AI.
A Leap in Sign Language Translation
These new models succeed Talksign's initial foundation model, representing a categorical leap from research curiosity to a deployable product. The company aims to address the complex challenge of teaching machines to understand sign language, a multi-dimensional system involving hands, facial expressions, and body movement. This launch signifies a major advancement in making AI-powered accessibility a practical reality for everyday use.
Palm 1.0 functions as the company's sign-to-text engine, interpreting ASL with remarkable precision and contextual understanding. The model achieves 84.2% semantic accuracy, ensuring it correctly captures the intended meaning behind a signer's gestures. This high level of performance, combined with 79.6% word-level accuracy, brings the technology into a range suitable for real-world deployment.
The model's accuracy is powered by a proprietary system called a Spatial Attention Graph Encoder, or SAGE. This architecture tracks 133 anatomical landmarks across the body, analyzing them as a connected graph rather than as isolated points. This holistic approach allows Palm 1.0 to interpret signing as a continuous, contextual conversation, much like a human would.
Generating Photorealistic Communication
Complementing this is Echo 1.0, a model that converts written or spoken language into photorealistic ASL video. It generates high-fidelity avatars signing at 30 frames per second with a translation latency of just under 30 milliseconds. This real-time performance is crucial for facilitating natural, conversational interactions without the perceptible delay found in older systems.
Unlike simpler tools that transcribe word-for-word, Echo 1.0 translates English into ASL gloss, preserving the unique grammar and syntax of sign language. This linguistic accuracy is critical for fluent and clear communication, avoiding common translation pitfalls. The system also supports personalized avatars generated from a single reference photo, enhancing the natural feel of the interaction.
Real-World Impact and Future Roadmap
Talksign intends for these models to serve in contexts where professional interpreters are scarce, such as in hospitals, classrooms, or for emergency alerts. The company emphasizes its collaboration with the Deaf community, ensuring datasets were reviewed by advisors and educators. User privacy is also a key consideration, with landmark extraction occurring on-device to protect raw video data.
The company acknowledges current limitations, including Echo 1.0's initial support for English only and the need for fine-tuning for specialized vocabularies. Future updates plan to incorporate additional languages like Spanish and French, as well as other sign languages like British and Nigerian. The goal is to continually expand the models' capabilities and accessibility across different regions and professional domains.
The introduction of Palm 1.0 and Echo 1.0 represents a significant step forward in assistive technology for the Deaf community. By moving beyond research and into production-ready systems, Talksign is paving the way for a more inclusive and accessible digital world. The full platform is scheduled for public release on mobile, desktop, and smart glasses by August 20, 2026.