After photos, Google targets video enhancements with on-device AI: Report

Reportedly, the Google Tensor G5 will support "Video Generative ML" features for AI-powered video editing. The chip is expected to debut next year in the Pixel 10 series

Google is reportedly planning to introduce significant software upgrades and enhanced artificial intelligence features for its future Pixel smartphone models with subsequent generations of its Tensor chips.

 

Consumer technology news platform Android Authority has reported that it has accessed several internal documents from Google’s gChips division detailing new AI features and upgrades coming to the Pixel 10 and Pixel 11 series smartphones with the next generation of Tensor chips. This includes a generative video editing feature, additional photo editing tools, improved camera optimisation, and more. Here is what it found in the documents:

 

Tensor G5: Focus on on-device AI for videos

 

 

The report states that with the Tensor G5 chip, expected to debut next year with the Pixel 10 series, Google will introduce “Video Generative ML” feature, which is described in the document as “Post-capture Generative AI-based Intuitive Video Editing for the Photos app.” The mention in the document suggests that the feature would function as a Magic Editor-like video editing tool within Google Photos. Additionally, Google is likely to incorporate this feature into the YouTube app for YouTube Shorts.

Furthermore, Google is said to be developing several new AI-powered picture editing tools, including “Speak-to-Tweak” and “Magic Mirror,” although not much is known about the functionality of these features. There is also a “Sketch-to-Image” feature anticipated to be similar to what Samsung offers with its Galaxy AI.

 

The report also notes that Google’s next Tensor chip will be capable of running Stable Diffusion-based models on-device, allowing apps like the Pixel Studio app to generate images locally instead of relying on cloud processing.

 

In terms of camera-related functions, the upcoming Tensor G5 chip could support 4K HDR recording at 60fps. In comparison, the current generation Tensor only supports up to 4K 30fps HDR video.

 

According to the report, Google plans to enhance image processing capabilities with the Pixel 11 series, utilising Machine Learning for 100x zoom in both photos and videos. The internal documents also mention a “next-gen” telephoto camera, suggesting that advanced zoom capabilities will be achieved through hardware improvements. The Pixel 11 is also expected to enhance Cinematic Blur and Night Sight video features.