Copy link to clipboard
Copied
I wamted to use the same generative fill background that was created on a few different images in a series. However, you can't just drag and drop, then mask on the additional images as the layer will also have parts of the previous photo in it, instead of just the images pulled by Generative Fill AI. Also when using the same prompts on the next image, the results are different. is there a way to get consistent results in a series of photos?
Colin Smith has covered this in one of his Photoshop Cafe videos.
Copy link to clipboard
Copied
Colin Smith has covered this in one of his Photoshop Cafe videos.
Copy link to clipboard
Copied
This is a good tutorial, however, I don't believe it answers the original question. If you have a series of photos of an individual (say 5) all with the same background and editing. You're able to generate an AI background for the first one, but the 4 remaining pictures with have different backgrounds. That's the issue at hand IMO.
Copy link to clipboard
Copied
Its a big problem if you can't have the same background on a series of images. You can try and describe the background but it is pot luck if they look similar.
Copy link to clipboard
Copied
Did you watch the video @Trevor.Dennis posted a link to?
Copy link to clipboard
Copied
The video doesn't answer the question. If you take a set of photographs of a car taken in the same location from different angles, it would produce a different background for each different shot. The background that it produces for the rear angle of the car would be different from the front angle taken of the car.
Copy link to clipboard
Copied
You made no mention of changing perspectives in your first post on this thread.
It sounds like you expect Genarative Fill to effectively create and store 3D environments and I don’t think that’s plausible for the time being.
Maybe you should pop over to the Photoshop (Beta) Forum and post a Feature Request.
Copy link to clipboard
Copied
I agree, unfortunately the tutorial does not answer to the original question. I am not yet sure if this problem can be solved using a script.
Copy link to clipboard
Copied
Agreed. Interesting and useful videos but....
Is there a way to save a result of generative fill as a preset or a style like you can with regular layer effects and filters? I'm fine with generative fill giving me 3 samples to choose from but what if i really love one of those and want to apply the EXACT result/output across other images?
Copy link to clipboard
Copied
You can save your preferred results as images/Patterns/in the Library/… for reuse, but If you want Generative Fill to behave differently this thread is not the ideal place to voice your wishes.
Post a Feature Request (»Idea«) or post over on the Photoshop (Beta) Forum.
Copy link to clipboard
Copied
I agree with Original Poster, for example I have 5 photos exactly the same aside from a model pose change Gen Fill provides totally different result when i resize the photos, meaning it's almost impossible to have an identicle set.
The background is 100% the same, but it does not provide any same results
Copy link to clipboard
Copied
The background is 100% the same, but it does not provide any same results
What makes you think it should be?
Have you followed the recent coverage about generative AI?
If the images are »identical« in background and perspective you can use the same newly created background for all of them, even as a linked Smart Object so that possible changes will apply to all of them.
Copy link to clipboard
Copied
I also want to do this. I have 5 photos of a woman on a wood floor and I want to replace it with the same concrete floor but when I generative fill every image gets a different style or colour of concrete floor.
Copy link to clipboard
Copied
Please provide a better description.
Posting screenshots can help oh so much when talking about images …
If the color of the replacement elements is the only problem then Photoshop offers several options to address that via Adjustment Layers.
Copy link to clipboard
Copied
None of the above answer the question....which I have posed before. As an example: I am a pet photographer and want to do holiday images using the same backdrop created in PS generative fill. The exact same prompts creates a distinctly different backdrop for each dog, making it difficult to offer clients a specific backdrop. I have also used Midjourney to create beautiful backdrops...but when you add any animal...Midjourney does not create the exact animal you shot....and unless a client wants a 'different' version of their pup, they will be far from pleased. Compositing pups into AI backdrops is far from ideal and it usually looks like the dog is simply 'floating' on the backdrop. I am at my wits end trying to get a decent composite of a dog on a digital backdrop...and would appreciate any suggestions for tutorials. OH...and does anyone know if generative fill will ever be able to create the same backdrop with the same prompts?
Copy link to clipboard
Copied
Compositing pups into AI backdrops is far from ideal and it usually looks like the dog is simply 'floating' on the backdrop.
Wouldn’t that be connected to your compositing skills? (edit: Compositing of different images/image parts is a task for which Photoshop has been used for a long time and it is not dependent on AI per se.)
Could you post examples (the original photograph, the AI-created background and your result)?
Copy link to clipboard
Copied
Hopefully this will explain what they are asking so you will understand it. You have 3 pictures of a ball. You use generative ai on the background of the first picture of the ball. You type in "green grass". So now you have a picture of a ball on a green grass background photoshop created via generative fill. You now want to use that exact same background on the other two ball pictures. If you use the exact same words for the prompt, green grass, but it creates a different background than what was generated for the first ball. How do you make generative fill generate the exact same background for all three ball pictures? It's not a matter of composite skills. It's a matter of replicating the background across different pictures. You can use the exact same prompt and not get the same result.
Copy link to clipboard
Copied
If you have the result you want then use that in the other images instead of trying to use Generative Fill again and again.
If you expect Generative Fill to create the same »scene« in multiple different 3D views you may want to check if such a feature has ever been announced by Adobe personnel and feel free to post a Feature Request.
Copy link to clipboard
Copied
I think it would make sense for Adobe to make this happen and I have already made a Feature Request.
Copy link to clipboard
Copied
Then let’s hope some Adobe employee will chime in.
Maybe post the link to the Feature Request here, too.
Copy link to clipboard
Copied
I agree this is a problem with getting the same AI generated background on multiple images.
One way that worked for me (I am creating the same backdrop on multiple portraits) is to recreate one clean background image without any subject (in fact I used AI to fill the gap where the subject used to be) and then I used that background behind all the portraits. Getting the subjects not to look floating will require some compositing. You could also create shadows using AI...
Copy link to clipboard
Copied
Clipping an object in an image and compositing it with another background, no matter how this was created, presents its own challenges.
Photoshop’s Selection Tools have been improved in recent years but for some/many cases they may not provide a very good results without some additional, manual processing.
Copy link to clipboard
Copied
I've been playing with Neural filters (harmonization). It helps some with color matching. But blending requires a lot of manual work.
Copy link to clipboard
Copied
We should be able to select from the results. And have a "reuse this fill" option. Then the next frames will be consistent. Adobe please give us this feature. Thanks.
Copy link to clipboard
Copied
100% agree.
if the AI creates something desirable, why not have the ability to apply that process repeatedly?