Google’s artificial intelligence (AI) company DeepMind has announced it will be making a new sign-language-to-text translation model (SL2T) available on its Pixel 11 phone, turning American Sign Language (ASL) into English.
The technology, offered through the Gboard keyboard and Live Transcribe, allows users to sign instead of typing when searching the web, creating messages or documents, interacting with the Gemini AI software and in conversations.
SL2T has been trained in over 50 sign languages and on more than 100,000 hours of data – of which around a quarter is in ASL.
Instead of using a camera feed, SL2T uses “pose landmark locations” and geometric coordinates for its translations, immediately discarding the original video and bypassing annotations known as “glosses” which Google says fail to “capture rich, non-linear aspects of sign language such as non-manual markers and spatial constructions”.
The company also acknowledges “occasional errors” with the tech, around rare signs, rapid fingerspelling, passive constructions, classifier depictions and tense without context (such as “kicked off” instead of “start”).
In a news article about SL2T, published on Wednesday, Google DeepMind’s sign language team said they “believe in building with the Deaf community, not just for it”.
“Deaf perspectives have shaped every stage of this project — from conceptualization by Sam Sepah, a Deaf Googler, to data collection with Deaf partners, evaluation in Deaf user studies, and impact assessment of the technology with Deaf experts.
“To guide responsible real-world deployment, we established the AI Sign Language Advisory Committee (AISLAC), bringing together many global Deaf organizations and subject-matter experts.
“Through this participatory governance model, the communities most impacted by our technology directly influence our development priorities,” they said.
The team also said more devices would offer the technology soon, with additional sign languages to follow.
News of SL2T comes more than a year after Google DeepMind announced the launch of SignGemma, which it described as its “most capable model for translating sign language into spoken text” that would launch later in 2025.
However, speaking to DeafEXPO attendees in Birmingham in May, Google’s EMEA (Europe, Middle East and Africa) lead for accessibility and disability inclusion, Christopher Patnoe, said the company hadn’t hit its “quality metrics” for the technology.
“So we’re going to continue to work on this with deaf people, or partner with organisations literally around the world, and we’re going to share it when we’re ready, starting with ASL, then different languages.
“This is coming. We want to make sure it’s good enough again,” he said.
It is not clear if SL2T is the new name for SignGemma.
Google has been approached by Liam O’Dell for comment.
Images: Google DeepMind.