FLUX.1-dev ControlNet Upscaler: Turn Blurry Images into Clear Visuals

I was digging through ComfyUI workflows the other day, trying to salvage some old, blurry photos—you know the ones. Grainy, slightly out-of-focus shots that just never turned out right. Then I stumbled on the FLUX.1-dev ControlNet Upscaler, and honestly? It worked better than I expected.

Here’s the thing: this model was trained on a ton of messed-up images—blurry ones, noisy ones, even those weird JPEG artifacts we’ve all dealt with. The idea is simple: it learns from the bad stuff, then cleans up your images. I gave it a shot, and the results were surprisingly solid.

It’s a ComfyUI tool that takes blurry or low-quality images and sharpens them up. Not in that overprocessed, AI-weirdness way—just cleaner, more defined details. I tested it on a few photos where the focus was just slightly off, and it actually pulled back some of the lost clarity.

If you’ve ever tried to upscale something before, you know most tools either oversharpen or leave it looking artificial. This one’s different. It keeps things natural while fixing the worst of the blur.

I grabbed the ControlNet model from Jasper’s Hugging Face repo and the main FLUX.1-dev model from Black Forest Labs. Dropped them into ComfyUI, wired up a quick workflow, and ran a test. No fancy tweaks—just the basics.

I didn’t expect this part, but it handled subtle motion blur way better than I thought it would. Not perfect, but definitely usable. For reference, I was working with some old DSLR shots where the autofocus missed slightly. The upscaler didn’t magically fix everything, but it made them look intentionally soft instead of accidentally blurry.

Anyway, if you’ve got a folder of “almost good” images lying around, this might be worth a try. I’ll dig into the settings more in the next part—there’s some interesting stuff with the ControlNet adjustments—but even out of the box, it’s pretty capable.