Generative ai

[jay fresno]

I use Edit > Generative AI to fill in areas around the edges of an image or to remove unwanted items. I do not use it to add objects that were not in the photo to begin with.

The resolution of the AI fill is very poor. Is there a way to get better resolution patches when using Generative AI?

[AlanGilbertson · Accepted solution]

As a general rule, keep any additions below 2k x 2k pixels for best image quality. Smaller is better. Firefly isn't set up to work with larger pixel dimensions. If you try to fill areas larger than that, the quality goes down very sharply.

Use the tools for how they are designed. If you want to remove something, use the Remove tool in preference to Generative Fill. In Ps 2025, the Remove it can use genAI if it needs to (check the options in the Control bar), but won't if it doesn't, and it's very effective. If you need to replace something in the image with something else, first remove it, then use Generative Fill with the appropriate prompt to add back what you need.

If you want to add to the edges of an image, you can use Generative Expand, which is an option of the Crop tool, but again keep any genAI to increments below 2048 px in any dimension.

You can use Generative Fill, which appears automatically in the Properties panel whenever you make a selection, to extend the boundaries of an image larger than 2048 px on that edge. First extend the canvas, then marquee select sections and fill them.

[jay fresno]

Alan, Thank you for your thorough reply. 

[AlanGilbertson]

Sure, happy to help. The best way to learn the strengths and limitations of genAI in Photoshop is to play with it. Start with a generated image, maybe, directly from Firefly; or an image that's been in your collection forever, waiting for the day it became feasible to remove all the distracting wires, Coke cans, and McDonalds wrappers messing up the scenic sunset. The wrong time to be exploring new features is when you're on deadline. 😉

[ahsanali123]

Hey Jay, great question — and I totally feel you on the resolution thing.

Right now, Generative AI tools (like the one in Photoshop or Firefly) often prioritize speed and blending over resolution, especially when filling large areas or complex patterns. But here are a few things you can try to get crisper results:

:hammer_and_wrench: Tips for Higher-Quality AI Fills:

1. Work on a higher-res image – Upscale your image before editing if possible. Some tools perform better when starting with more pixels to work with.

2. Use smaller selection areas – Instead of filling a large chunk at once, break it down into smaller selections. This gives the AI a better chance to generate high-quality, detailed patches.

3. Refine edges manually – After the AI fill, use tools like the Smudge, Clone Stamp, or Healing Brush to clean up blurry or odd edges.

4. Layer-based approach – Do the AI fill on a separate layer. You can then mask and blend only the parts you need, keeping your original detail intact.

Hopefully Adobe will continue to improve resolution handling in future updates — it’s a hot topic in the creative community for sure.

By the way, I’ve been exploring some cool AI image tools myself while working on word to ppt ai  — super fun (and challenging) stuff! Always great to connect with fellow AI creatives 🙌

Let me know if you find a workaround that works well for you!

[Nick_Ponomarev]

I'm really wondering why Adobe's artificial intelligence is so poor. Why isn't it like Midjourney's? I thought Adobe was a richer company, but its AI feels like it's stuck at the level of Stable Diffusion version 0.9. Adobe, you need to do something about this, or you'll end up at the tail end of innovation. Midjourney already has a video generator, and DALL·E is producing decent results. Catch up already, or your competitors will crush you.