Galaxy S25’s on-device AI capability expands, reducing cloud dependency

Moving AI processing to powerful cloud data centers enables impressive features and could promote good smartphone battery life. It also introduces lag and requires an active internet connection when using features like Samsung Notes’ Summarize or Samsung Gallery’s Generative Editing.

As diligent digital sleuth AssembleDebug discovered on X/Twitter, the Galaxy S25 family will start to change that by processing AI generative photo editing locally on your device (Source). This should further help detach daily activities from the cloud using the Snapdragon 8 Elite’s vast capabilities.

Putting advanced AI in the palm of your hand

The start of what is (hopefully) a move to on-device processing

While examining the code underlying Samsung’s versatile collection of apps, prolific leaker and feature explorer AssembleDebug spotted a set of comments describing some features available for different chipsets. Among them was a reference to the SM8750, Qualcomm’s popular Snapdragon 8 Elite system-on-chip. The code comments name three tools apparently related to the upcoming S25 series:

• FEATURE_WALLPAPER
• FEATURE_INOUT_PAINTING
• FEATURE_GEN_EDIT_ON_DEVICE

The last of the features offered by AssembleDebug found that sports was an obvious enough name to deduce what was coming. Barring a wildly misleading naming scheme, “FEATURE_GEN_EDIT_ON_DEVICE” should enable AI generative photo editing on the Galaxy S25 series without the need to send images to the cloud. It’s a win for speed, convenience, and privacy.

The code also mentions inpainting and outpainting, which refer to mixing changes in an image and expanding an image outside of its original boundaries, respectively. In theory, these would fall under generative AI editing, although the minimal evidence so far does not explicitly indicate where cloud processing becomes a requirement. Increasing flexibility via on-device processing, and generally beefing up and improving Galaxy AI performance, could go a long way toward making this a viable reason to upgrade to a new Samsung flagship.

Driving adoption of on-device AI processing

An important step in the right direction

Currently, Galaxy AI generative photo lets you almost magically isolate objects in images and move, resize, or erase them, but cloud processing takes care of most of the work. While theoretically good for your battery life, it also forces your phone to upload all relevant data to the cloud and locks you out of resource-intensive features if you don’t have internet access.

Google’s Gemini Live toolkit, for example, loses almost all of its functionality when you’re offline. Galaxy AI takes a more localized approach in some cases, with translation tools in Samsung’s Phone, Keyboard and Notes apps, as well as Instant Slow-Mo image interpolation, already running exclusively on the device.

It’s unclear whether Samsung’s Sketch to Image feature will be sufficient for on-device processing.

The slow but significant migration to on-device processing is partly at the mercy of hardware. The powerful and flexible Snapdragon 8 Elite breaks new ground in enabling mobile devices to run demanding software. In terms of AI, the chipset allows wearable devices to operate data-intensive models quickly and without consuming excessive battery. The convenience and privacy of offline processing should help streamline AI tools and adoption. Only time will tell what additional new features Galaxy AI will bring to smartphones, but genAI photo editing likely won’t be the last.