So I actually tried running SD locally on an Android phone, and it's way more usable than I expected.

I always thought generating AI images had to rely on cloud GPUs, but I just found out Android phones can run Stable Diffusion fully locally. This open-source app called Off Grid, with NPU acceleration on Snapdragon flagships, can generate an image in five to ten seconds—no subscription, no internet needed, and your prompt never gets sent to any server.

The principle is the same: starting from random noise and denoising step by step, each step getting closer to the scene you described. It even gives you a real-time preview so you can watch the image form, no staring at a blank screen. Default is 512x512, 20 steps. CPU works too but slower, 30 to 60 seconds per image.

Sometimes when I’m doing UI work and need a few placeholder illustrations or inspiration images, I just pull out my phone and generate them without booting up my PC—that’s a game changer for me. Sure, the detail can’t match desktop, but it’s good enough for emergencies.

The fact that prompts don’t get uploaded to the server is a huge win for privacy nerds.

The preview updates in real time as the image generates, the experience is maxed out.

Snapdragon NPU does one image in 5 seconds, faster than my old GPU.

Just casually generating placeholder illustrations on the fly—this is the joy of being a designer.

512 resolution is still kinda limiting.

CPU takes 60 seconds? Might as well just wait for the computer.