Talksign, a Nigeria- and UK-based artificial intelligence startup, has launched two new AI models designed to enable real-time communication between American Sign Language (ASL) and spoken or written language.
The new systems Palm 1.0 and Echo 1.0 aim to close long-standing accessibility gaps by translating sign language into text or speech, and converting spoken or written language into lifelike sign language video.
Palm 1.0 interprets ASL into text or speech with 84.2% semantic accuracy, while Echo 1.0 performs the reverse, generating photorealistic ASL video from text or voice input using real-time avatars.
The launch, announced on May 20, builds on Talksign’s earlier foundation model, Talksign-1, introduced in February. That system could translate 250 ASL signs into text or speech but was limited to isolated signs and struggled with full sentences and fingerspelling limitations the new models are designed to overcome.
The rollout comes amid growing global attention on accessibility and communication barriers. The World Health Organisation estimates that more than 430 million people live with disabling hearing loss, with millions relying on sign language as their primary means of communication.
Despite this, many everyday digital tools still prioritise spoken and written language, limiting access for Deaf users across education, public services, and digital platforms.
Palm 1.0 is Talksign’s sign-to-text model, built to interpret ASL in real time. The company says it achieves 84.2% semantic accuracy and 79.6% word-level accuracy, allowing it to capture meaning from continuous signing rather than isolated gestures.
The model was trained on more than 71,000 ASL samples and uses a transformer-based architecture with a system called SAGE (Spatial Attention Graph Encoder), which tracks 133 body landmarks including hands, head, and shoulders to interpret movement in context.
“Palm 1.0 is the first model we are confident putting into the hands of Deaf users at scale,” said Edidiong Ekong, CEO and co-founder of Talksign. “The next step is putting it everywhere a Deaf person needs to communicate: on phones, smart glasses, in classrooms and hospitals.”
Echo 1.0, the company’s sign-generation model, converts text or speech into ASL video using photorealistic avatars. It was trained on 94,410 ASL sentence pairs and refined through multiple training cycles to improve fluency and accuracy.
Talksign says the system can generate video at 30 frames per second with a latency of about 29 milliseconds, making the signing appear near-instant in real time.
Unlike word-for-word translation, Echo 1.0 converts English into ASL gloss, preserving the grammar and structure of sign language before rendering it into 3D motion sequences. Users can also generate personalised avatars from a single photo, a feature designed to make communication feel more natural and familiar.
The system processes translation in three stages: speech recognition or text input, conversion into ASL gloss, and final rendering into animated sign language video.
Co-founder and CTO Kazi Mahathir Rahman said the technology moves sign language closer to becoming a “first-class interface” for digital interaction, especially for Deaf users engaging with AI systems.
Talksign says the models are intended for real-world use cases such as classrooms, healthcare, emergency alerts, and public broadcasting especially in situations where human interpreters are not available.
However, the company acknowledges current limitations. Echo 1.0 currently supports only English, with additional languages such as Spanish, French, and Arabic planned. It also requires further development for specialised fields like medicine and law, while Palm 1.0 still has constraints in fully capturing complex continuous signing.
Talksign plans to expand support to other sign languages, including British Sign Language, German Sign Language, and Nigerian Sign Language in future updates.
Full deployment across desktop apps and Meta Ray-Ban smart glasses is scheduled for August 2026.
The startup joins a growing field of AI-driven accessibility tools, alongside platforms such as SignVrse, which is developing real-time sign language translation using 3D avatar technology to improve communication for Deaf users globally.