Generative ai is a great way to remove anything from an image in Photoshop. Sometimes they don’t work well. You try to remove something, instead it’s replaced with something else, or the results leave a broken mess, or the resolution is really blocky and soft. If any of this is happening to you, this tutorial will help you by showing you,
- When to Use Generative Fill vs other tools.
- Why the remove tool works on some images and not others.
- Why Generative Fill doesn’t remove things, but replaces them with other objects instead.
- Why the resolution doesn’t look good, but why it doesn’t “ALWAYS” matter.
The tutorial
Should I use Generative Fill, or the Remove tool to remove this object? Here is a tutorial that explains the Remove tool, Generative Fill and the difference
You could use either. Depending on your needs, one is better than the other. Lets have a look.
The Remove tool in Photoshop
The Remove tool has the following advantages;
- No resolution limits.
- Its faster.
- It’s a local ai, so it doesn’t need an internet connection.
- It doesn’t use up your Generative Credits
First, we will remove the woman AND the bike. Choose the Remove tool from the toolbox.
Choose Sample All Layers from the toolbar to use the tool non destructively on a new layer.
Turn on Remove after each stroke.
Create a new layer in the Layers panel.
Paint over the area that you want to remove with the remove tool.
When you release, the remove Tool will remove the painted area.
It did a good job.
Why the Remove tool worked.
Look at the area we desire to remove.
Similar image areas exist in the image. Enough for The Remove tool to use to generate the new pixels.
Use the Remove tool to remove just the woman.
This time, paint over just the woman.
Notice it does not do a good job this time.
Why the Remove Tool failed.
Here is the area we want to remove.
There aren’t similar objects like the bicycle Frame in the image for the remove tool to use.
So let’s look at the disadvantages of the remove tool.
It can’t do complex generation (It can generate, to a certain extent, but not as good as Generative Fill).
There are no alternative variations to choose from.
Using Generative Fill to Remove an object and why it doesn’t always remove
Let’s take the same image and use Generative Fill. I’ll start by using a method, that some people assume is a good way to work, but often leads to failure. I’ll explain why and then show you the correct way to get the best result.
Choose the Object Selection tool from the toolbar.
Turn on Object Finder.
Roll over the image and click to select the woman. Alternatively, click Select Subject, or use your favorite selection tool. This isn’t a tut on making selections, how to make selections in Photoshop if you need a primer.
See the marching ants selection. We need to expand this, so we aren’t selecting part of the edges.
Choose, Select>Modify>Expand.
Choose 20 pixels.
Choose Generative Fill from the taskbar. Leave the text field blank and click Generate.
Instead of removing the person, it generated a different person. Why? Read on.
How to make Generative Fill Remove an object and not generate a new one
I know many people try to force Generative fill to remove an object by typing a command into the text fried such as “Remove Person”, and they swear it works. This is purely coincidence or at best a placebo tablet. The reason for this is that Adobe has said that instructions prompts don’t do anything. If you want to remove something, leave the field blank and click generate. Ok Colin. So why doesn’t it work?
There are 2 big reasons.
#1 You didn’t entirely select the entire object, make sure every part of the object is included in the selection, and make it loose, not a tight selection like you are used to.
Here is a biggie! Because we used a selection tool, look at the shape of the selection. It’s a prefect silhouette of a person.
If you remember, about a year ago I discovered something about Generative Fill and made a video on it
(It has since been copied without credit, by certain YouTubers, like this tip probably will be (July 13, 2024)
The shape of the selection is the MOST important thing when using Generative Fill.
Instead, use the lasso tool and make a loose and rough selection around your object.
Now when we apply Generative Fill, it perfectly removes our person, because it doesn’t think we are trying to make a person.
Resolution, it matters, but not as much as some people think
Ok, I hear this: “Generative Fill generates in low resolution, so it’s useless to me.” This is true and false at the same time. Let me explain.
If we look at our image and make a 1024 x 1024 pixel box, you can see the size that can be generated in Full resolution. This is the size that Gen Fill uses.. but..
What matters is the target size, not the source size. Please stick with me here and read the rest.
Lets make a larger selection and fill it with Generative Fill
If we go close up, we can see the resolution mismatch.
If you are using Photoshop Beta, click the Enhance button in the Options panel to use Super Resolution.
But, lets choose File>Export>Save For Web
Most people (not all, stay with me) are outputting for social media or electronic output. In this case, you would rarely go larger than 1920 on the large size.
If you look at 100%, at this size you cannot see the resolution mismatch. Some people are saying that Generative Fill doesn’t look good on any image larger than 1024.
The truth is, the patch area shouldn’t be larger than 1024px in the OUTPUT image.
Yes, if you are working on large images, like my Sony a1, which shoots at 50MP and you are outputting high resolution digital files or you are printing, then yes this is a problem. If you are sharing on social media and working in Low res
less than Full HD this doesn’t matter.
If you are working on larger files.
I acknowledge that some people are working in high resolution workflows, for a large part of my career that’s all I worked on. Try this:
1. Generate in smaller pieces. I originally shared this tut and provided an action a bit over a year ago, and recently updated that tutorial here.
The other thing you could try is to combine the techniques. For the example above, apply the remove tool and other tools on a layer. For the more detailed work, make a smaller selection and use Generative Fill. Now, use Layer masks and combine the methods for better results. Sounds like work? It’s still less work than doing content-aware fill and then having to clone the missing object, composite it from another photo or paint it in by hand, like many of us have done for years.
I hope you found this tutorial useful and it helps make things a bit easier for you
Colin
PS Don’t forget to follow us on Social Media for more tips.. (I’ve been posting some fun Instagram and Facebook Stories lately)
You can get my free Layer Blending modes ebook along with dozens of exclusive Photoshop Goodies here 
