Pixel 8 Pro runs Google’s generative AI models on-device

Google’s newly announced Pixel 8 Pro will be the first hardware to run Google’s generative AI models on-device, according to Rick Osterloh, SVP of devices and services at Google.

Onstage at an event today, Osterloh said that the Pixel 8 Pro’s custom-built Tensor G3 chip, which is designed to accelerate AI workloads, can run “distilled” versions of Google’s text- and image-generating models to power a range of applications, like image editing.

“We’ve worked closely with our research teams across Google to take advantage of their most advanced foundation models and distill them into a version efficient enough to run on our flagship Pixel,” Osterloh said.

Thanks to the on-device models, Google’s managed to improve Magic Eraser, its post-processing tool for touching up photos, so that it can remove larger objects and people smudge-free. This enhanced Magic Eraser generates new pixels to fill in the spaces left by anything removed from a shot, Osterloh says, resulting in a higher-quality finished image.

Zoom will get better, too, Osterloh claims, thanks to a new on-device model that can “intelligently” sharpen and enhance the details of photos.

And the benefits of on-device processing extend to audio recording. Soon, the Pixel 8 Pro’s recording app will deliver summaries of recordings that recap the highlights of meetings.

Elsewhere, a large language model running on the Pixel 8 Pro will power smart replies in Google’s keyboard app, Gboard. Osterloh says that the upgraded Gboard will generate “higher-quality” reply suggestions with better overall conversational awareness.

With the exception of Magic Eraser, which is available on the Pixel 8 Pro at launch, the on-device generative AI features will arrive in December via an update, Osterloh said.