Google has unveiled SignGemma, a cutting-edge artificial intelligence (AI) model designed to convert sign language into spoken text. This new model, a part of the Gemma series, is currently undergoing testing at the company’s Mountain View headquarters and is slated for launch later this year. Like its counterparts in the Gemma series, SignGemma will be an open-source AI model, making it accessible to both individuals and businesses. The model made its debut at the Google I/O 2025 keynote, with the aim of facilitating communication for individuals with speech and hearing impairments, even when conversing with those unfamiliar with sign language.
SignGemma Can Track Hand Movements and Facial Expressions
In a recent post on X (formerly Twitter), Google DeepMind’s official account shared a demonstration of SignGemma along with insights into its anticipated release date. The AI model was previously highlighted during the Google I/O event by Gus Martin, the Gemma Product Manager at DeepMind.
We’re thrilled to announce SignGemma, our most capable model for translating sign language into spoken text. 🧏
This open model is coming to the Gemma model family later this year, opening up new possibilities for inclusive tech.
Share your feedback and interest in early… pic.twitter.com/NhL9G5Y8tA
— Google DeepMind (@GoogleDeepMind) May 27, 2025
During the event, Martin emphasized that the AI model excels in delivering real-time text translations from sign language, thereby enhancing face-to-face interactions. SignGemma has been trained on datasets representing various sign language styles, with a particular proficiency in American Sign Language (ASL) when translating into English.
According to MultiLingual, SignGemma’s open-source nature allows it to function offline, making it an ideal choice for use in areas with poor connectivity. The model is built on the Gemini Nano framework and employs a vision transformer to monitor and interpret hand movements, shapes, and facial expressions. Google may also incorporate this model into its existing AI tools, including Gemini Live, while making it available for developers.
Describing SignGemma as “our most capable model for translating sign language into spoken text,” DeepMind confirmed that the model is set for a release later this year. Currently in its testing phase, the tech company has created an interest form to encourage individuals to participate in early trials and provide feedback.
