(1).png)
Google has introduced SignGemma, a new artificial intelligence (AI) model that can convert sign language to spoken text. The model, which will be part of the Gemma series, is now being tested by the Mountain View-based tech giant and is set to be released later this year. SignGemma, like the other Gemma models, will be an open-source AI model available to both individuals and enterprises. It was initially shown during the Google I/O 2025 keynote and is intended to assist persons with speech and hearing problems in efficiently communicating with those who do not understand sign language.
SignGemma can detect hand movements and facial expressions
In a post on X (previously known as Twitter), Google DeepMind's official account revealed a demo of the AI model as well as some information about its release date. However, this is not the first time we've seen SignGemma. Gus Martin, Gemma Product Manager at DeepMind, also gave a brief presentation during the Google I/O event.
We're thrilled to announce SignGemma, our most capable model for translating sign language into spoken text. 🧏
— Google DeepMind (@GoogleDeepMind) May 27, 2025
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
During the demonstration, Martins stated that the AI model can provide real-time text translation from sign language, allowing for seamless face-to-face conversation. The model was also trained on a variety of sign language datasets; however, it performs best when translating American Sign Language (ASL) into English.
According to MultiLingual, because it is an open-source model, SignGemma may work without a connection to the Internet. This makes it ideal for usage in places with poor connection. It is supposed to be based on the Gemini Nano architecture and use a vision transformer to detect and analyze hand gestures, forms, and facial expressions. Beyond making it available to developers, Google may include the model into its existing AI products, such as Gemini Live.
DeepMind described it as "our most capable model for translating sign language into spoken text," and announced that it will be published later this year. The accessibility-focused broad language model is presently in early testing, and the tech behemoth has released an interest form to encourage people to check it out and submit feedback.
