The low resolution of the generated stuff makes it mostly useless for me so far. I'm working on 61 MP images, getting ONE of those megapixels filled at a time totally wrecks the experience. And it takes forever. It really needs to match whatever resolution is necessary for the filling!
You can easily work around the resolution/details problem: instead of using generative expand tool, just expand the canvas, go to rectangular selection and make a selection that does not exceed 1024 on the larger side and be aware of overlapping a little the original image and the click on generate, then choose the best option. The repeat more times this step to cover all the missing area. Then select the original photo and all the generative layers you created, group them and create a smart object of the group. The continue your workflow. You can start from a 2000x3000 photo and continue to expand it to 6000x9000 without loosing any detail.
Yup, that is 100% true, and I will be creating another video walking through that workflow. It's not the most efficient way to use Generative Fill, but it certainly does address the current 1024px limit.
And unfortunately made almost useless by Adobe. as with the new credit system being implemented by Adobe, its gonna eat your pool of credits real fast to perform lots of repeted generative fills to create the expansion.
The different resolution is a real issue. From a distance things look fine, but when you zoom in on your picture you'll see a big difference. Besides that Photoshop still has errors when you're editing with generative fill. Out of the blue you'll get a warning that what you are doing does not comply with the user guidelines, however there is an easy way to avoid that by entering a dot in your prompt window.
I can see some uses of this technology in my photos, in removing unwanted objects. Going to be fun experimenting and thanks for help in getting us started on this new adventure.
That helps to see (for the expand around the bldg) that you didn't have to use a rectangular marquee to designate the empty areas to be filled! BTW, love the addition of the plane. It looks like how the planes look flying past downtown buildings (including Adobe's) in San Jose, CA.
Happy to help! Yeah, I prefer using the crop tool for any Generative Expand tasks because it makes things so much easier to visualize (for me, at least). Now if only they could increase the resolution of the fill content.
Thanks Brian. Good recap on what to do with a regular photo. I found I used Gen Expand a bunch in Beta since I don't seem to leave enough room sometimes. Used it once this morning and ran into the lower resolution and a bit of a line being visible. Posted the photo on Mastadon and I'm sure no one will ever notice. If I was printing maybe I'd take the time to do 1024 px section at a time. Thanks !! - Tim
Glad you enjoyed the video, Tim! You nailed it, too. The 1024px limitation is fine if you're going to share it on social media. As you know, hardly anyone scrutinizes photos in a social feed. It's more of a spectator sport, so I don't think anyone would notice the reduced resolution of filled areas. Still, it'll be great when Adobe does increase that.
Yup, that was very frustrating. The only way I was able to get a photo from LR to PS (not Beta) was to have Photoshop v24 already loaded, and then LR would send it to the correct version.
CC shows me only version 24.7 (updated today) not 25 & there is no generative component present. I do see a PS beta 25.1 which is new today also. I'm in the US. Wonder why I don't see what you see?
I asked this same question on Colin Smith's Photoshop Cafe channel earlier today: Does this mean were are not going to get a major version update in October during MAX? V25.0.0 is welcome - and appears to be stable judging by the lack of posts on the Adobe Photoshop Ecosystem forum - but there is nothing exactly new in it. It just consolidates features we have had in the beta version for quite a time now. I can kind of understand Adobe needing to prevent Midjourney and Dall-e steeling a march on them, but with 30 million Creative Cloud subscribers, that was never really going to be a problem. I sincerely hope we will get an October update with some brand new features. I have at least one fellow content creating professional who feels that Firefly is pointless. My own thought is that we have to embrace it and adapt to it, and there is still plenty of room to use skill and imagination to create better content than the average hobbyist. I am a big fan of Colin Smith and Jesús Ramirez and they have demonstrated workflows that will help people keep an edge. I have yet to watch Brian's video, but I am about to put that right, and I am sure he will have some useful tips to add to the debate.
Thanks, Brian. I watched the video with interest yesterday not expecting it would come in quite handy today. I was testing a couple of 50mm lenses on my 60-mp Leica 11 this afternoon and took shots of a little wooden bell tower on the outside of a church, aiming to test the lens resolution on the bell and forgetting about the apex of the tower. Once in LR, I realized it was a mistake nor include the top of the tower, which ends in a triangle. Bingo! I wonder if even the parishioners would recognize that what AI generated isn't the actual top of their tower. To be sure, the lack of resolution is apparent once I zoom in, so I'll be waiting for Adobe to address that.
No one seems to be concerned with the fact that the latest version has very different Requirements to run. Operating system Windows 10 64-bit (version 22H2) or later; LTSC versions are not supported
Thanks for the informative video - I enjoyed watching it. I have some reservations about using some of this in my own work, though. I photograph wildlife, so using AI for things like expanding the canvas and removing objects (branches etc in my case) is going to speed up my workflow. I've already been doing things like that for years because birds and animals aren't always cooperating! I try to get it right in camera, but if necessary I'll expand a canvas or remove a branch. I never add elements to my images that weren't there in the first place, though - and that's where my problem with AI begins because it makes it so easy and realistic to do this. I think it's where the nature of photography is going to shift from record to interpretation. It's always been both of these things - we take images of real life (record) and then improve on tone, colour and basic composition in the edit (interpretation). With AI, our "real life" image can become just the basis of something else - and that "something else" could have very little relationship to what we saw. It's already got to the stage where I don't trust the photographs that I see unless I absolutely know and trust the photographer. It's quite sad to think that in future, any stand-out genuine photograph is going to lead the viewer to ask "yes, but is this real or AI?". I guess AI is going to mean there'll be a lot more stand-out "images" around in the future, but probably fewer stand-out "photographs."
I’d be interested in this too. Will say I haven’t tried much with people because it’s not so good. But the other day I was trying to take a person completely out of a photo. PS wouldn’t do it. But it did replace with a less obtrusive person I could live with. And actually you can’t tell in less you pixel peep.
Why? The computational resources to perform these generative tasks is huge, especially when multiplied by the entire potential Creative Cloud subscriber base. Try using Midjourney. You can’t… not without paying for a plan that also limits you in some way.
Even though it will probably not happen, I truly hope one day when everything has been well-integrated, tied up and validated to their servers, Adobe will announce that they will charge users each time they launch Photoshop. Perhaps only then some of them will realize the dangerous direction we are heading on.
Let's say you have a complete photo set, how do you use it so that every photo looks like it's taken so that they "match" the style, light, color, and whatever else that keeps it looking like it was all done in the same shoot? Like who would want 50 photos with 50 different looking styles, light and whatever else.
Okay asking for pie in the sky i know but if Adobe AI understands that that is a building based on it's photo library and image recognition software, wouldn't it be great if it knew what building it was and, if it had other image references for that building, build and intelligent expansion with the environment that already exists at that location. I thought about this when i was playing around with a Seattle skyline photo that i have. the expansion was good but anybody who lives in Seattle would see that it was off a bit. I'm sure Adobe has city skylines in its library. Could they not then expand the skyline with buildings that exist within the skyline? Wouldn't' that be nice? Put that bug in your buddy's ear.😉
@@joelwexler Yes take our vacation photos from the privacy of your home office. I had a wonderful time BTW but my wife is wondering who that blonde in all my pictures is.
Let's not forget that the new updated AI features in PShop 25, even though they are now approved for commercial use, are Only, 1024 resolution! That's a major drawback for anyone shooting Raw on High-resolution cameras.
I read somewhere that after Christmas Adobe will be selling credits to use the generative fill, and for each time you press the generate button you’ll use some credit. I was very disappointed to here that. Have you heard any rumours?
I did read about the new credits system, and Adobe published more detailed info about it here: helpx.adobe.com/firefly/using/generative-credits.html It looks like credits go into effect for paying subscribers on Nov 1, 2023, and you can check your credit balance in the Creative Cloud app (after Nov 1). Also, it's important to note that credits don't carry over from month to month, so it's a "use or lose" situation. Still, this does not surprise me. Remember: performing these generative AI tasks requires a lot of computational power, and it can easily be abused without any sort of credit system. Midjourney has been doing this for a while now, and they don't even offer a free trial option anymore.
@@brianmatiash thanks that’s very helpful. It’s also not as bad as I thought. Photography plan gets 250 a month unless you’re subscribing after November.
Yeah, it’s pretty reasonable, especially because I don’t necessarily see photographers in particular blowing through their Gen AI credits each month. Graphic designers and digital artists? Possibly. But, I suspect that this tech has limited, albeit tremendously powerful, applications for photographers.
I have the complete cloud so I will get 1000 a month. More then enough. And CF will not cease to function if you run out, it will just become slower...how much slower is anyone's guess.
@@brianmatiash What a rip-off! I'm definitely only ever using it to remove things the eraser tool can't fix. Imagine the number of AI masks you could create in Lr worked that way - I'd immediately ditch LrC, forever.
Hmmm, I've only focused on the Generative AI tools so far, but will take a look at the remove tool in a bit. Although, I've never run into that issue before. The most basic first steps I'd recommend would be to uninstall and reinstall Photoshop. If it keeps happening, then Adobe Customer Support should be your next step. Let me know if the update fixes the issue for you.
Just tried it and it's working "OK" for me. I did notice that PS will fill in the whole area if say you are highlighting a bits of a fence to remove. Haven't gone any deeper on that.
I am finding that the images it produces even with minimalist text prompts are amazingly accurate to what I envisioned. I wouldn't be surprised to find that eventually the AI, if we have a subscription, will begin to learn what style of images we tend to like and give us ones akin to our preceding likes and dislikes. Seems like a logical progression to me. From a subscription loyalty, financial standpoint it would make me even more wedded to them and their tech.
Erik I have been using it as an alternative to hunting through stock images. It gets so close to what you see in your head if you describe it well enough, and it is way faster than searching stock libraries. The 1024 x 1024 resolution limit is a problem, but it's totally useful for finding (generating) smaller image elements. What makes it so doable is how amazingly good Photoshop's selection tools are now.
The low resolution of the generated stuff makes it mostly useless for me so far. I'm working on 61 MP images, getting ONE of those megapixels filled at a time totally wrecks the experience. And it takes forever. It really needs to match whatever resolution is necessary for the filling!
There are programs like Topaz Photo AI that have upscaling (& denoising & sharpening)... that work as plugins to Ps.
Once they upgrade that’s ITS A WRAP!$!
You can easily work around the resolution/details problem: instead of using generative expand tool, just expand the canvas, go to rectangular selection and make a selection that does not exceed 1024 on the larger side and be aware of overlapping a little the original image and the click on generate, then choose the best option. The repeat more times this step to cover all the missing area. Then select the original photo and all the generative layers you created, group them and create a smart object of the group. The continue your workflow. You can start from a 2000x3000 photo and continue to expand it to 6000x9000 without loosing any detail.
Yup, that is 100% true, and I will be creating another video walking through that workflow. It's not the most efficient way to use Generative Fill, but it certainly does address the current 1024px limit.
At 320AM that's amazing. I hope I remember.
Copied your text to desktop. Thanks
And unfortunately made almost useless by Adobe. as with the new credit system being implemented by Adobe, its gonna eat your pool of credits real fast to perform lots of repeted generative fills to create the expansion.
The different resolution is a real issue. From a distance things look fine, but when you zoom in on your picture you'll see a big difference. Besides that Photoshop still has errors when you're editing with generative fill. Out of the blue you'll get a warning that what you are doing does not comply with the user guidelines, however there is an easy way to avoid that by entering a dot in your prompt window.
Thanks Brian for the presentation
You're most welcome, Allison!
Thank u all very much
I can see some uses of this technology in my photos, in removing unwanted objects. Going to be fun experimenting and thanks for help in getting us started on this new adventure.
Thank you very much Brian for your explanation!!! Very didactic! Cheers
You’re most welcome! And didactic, too! I really appreciate it 😊
That helps to see (for the expand around the bldg) that you didn't have to use a rectangular marquee to designate the empty areas to be filled! BTW, love the addition of the plane. It looks like how the planes look flying past downtown buildings (including Adobe's) in San Jose, CA.
Happy to help! Yeah, I prefer using the crop tool for any Generative Expand tasks because it makes things so much easier to visualize (for me, at least). Now if only they could increase the resolution of the fill content.
Thanks Brian. Good recap on what to do with a regular photo. I found I used Gen Expand a bunch in Beta since I don't seem to leave enough room sometimes. Used it once this morning and ran into the lower resolution and a bit of a line being visible. Posted the photo on Mastadon and I'm sure no one will ever notice. If I was printing maybe I'd take the time to do 1024 px section at a time. Thanks !! - Tim
Glad you enjoyed the video, Tim! You nailed it, too. The 1024px limitation is fine if you're going to share it on social media. As you know, hardly anyone scrutinizes photos in a social feed. It's more of a spectator sport, so I don't think anyone would notice the reduced resolution of filled areas. Still, it'll be great when Adobe does increase that.
Thanks!
Thank you so much!
So glad that it is finally out of beta. My LR kept going to beta. I now have erased beta and going with PS 25.0!
Yup, that was very frustrating. The only way I was able to get a photo from LR to PS (not Beta) was to have Photoshop v24 already loaded, and then LR would send it to the correct version.
great video and tip updates on genfills
Glad you found the video helpful! Thanks for watching
VERY HELPFUL !
great, thank you
CC shows me only version 24.7 (updated today) not 25 & there is no generative component present. I do see a PS beta 25.1 which is new today also. I'm in the US. Wonder why I don't see what you see?
Well that's weird. Now I have it. Apparently CC doesn't check automatically for updates or it's infrequent.
Hi, Brian. Thanks for the video. Adobe could launch an upscaler, too.
Big news, helpful video! On Creative Cloud I am seeing two updates - Ps 24.7.1 and PS 25.0.
Do I need to update 24 before 25?
I wouldn't think so. You may have the option to keep both versions installed, which is why there are two updates.
@@brianmatiash Thanks, I see that now. I can keep the old until I am confident with the new release.
I asked this same question on Colin Smith's Photoshop Cafe channel earlier today: Does this mean were are not going to get a major version update in October during MAX? V25.0.0 is welcome - and appears to be stable judging by the lack of posts on the Adobe Photoshop Ecosystem forum - but there is nothing exactly new in it. It just consolidates features we have had in the beta version for quite a time now. I can kind of understand Adobe needing to prevent Midjourney and Dall-e steeling a march on them, but with 30 million Creative Cloud subscribers, that was never really going to be a problem. I sincerely hope we will get an October update with some brand new features. I have at least one fellow content creating professional who feels that Firefly is pointless. My own thought is that we have to embrace it and adapt to it, and there is still plenty of room to use skill and imagination to create better content than the average hobbyist. I am a big fan of Colin Smith and Jesús Ramirez and they have demonstrated workflows that will help people keep an edge. I have yet to watch Brian's video, but I am about to put that right, and I am sure he will have some useful tips to add to the debate.
Thanks, Brian. I watched the video with interest yesterday not expecting it would come in quite handy today. I was testing a couple of 50mm lenses on my 60-mp Leica 11 this afternoon and took shots of a little wooden bell tower on the outside of a church, aiming to test the lens resolution on the bell and forgetting about the apex of the tower. Once in LR, I realized it was a mistake nor include the top of the tower, which ends in a triangle. Bingo! I wonder if even the parishioners would recognize that what AI generated isn't the actual top of their tower. To be sure, the lack of resolution is apparent once I zoom in, so I'll be waiting for Adobe to address that.
No one seems to be concerned with the fact that the latest version has very different Requirements to run.
Operating system Windows 10 64-bit (version 22H2) or later; LTSC versions are not supported
Thanks for the informative video - I enjoyed watching it. I have some reservations about using some of this in my own work, though. I photograph wildlife, so using AI for things like expanding the canvas and removing objects (branches etc in my case) is going to speed up my workflow. I've already been doing things like that for years because birds and animals aren't always cooperating! I try to get it right in camera, but if necessary I'll expand a canvas or remove a branch. I never add elements to my images that weren't there in the first place, though - and that's where my problem with AI begins because it makes it so easy and realistic to do this. I think it's where the nature of photography is going to shift from record to interpretation. It's always been both of these things - we take images of real life (record) and then improve on tone, colour and basic composition in the edit (interpretation). With AI, our "real life" image can become just the basis of something else - and that "something else" could have very little relationship to what we saw. It's already got to the stage where I don't trust the photographs that I see unless I absolutely know and trust the photographer. It's quite sad to think that in future, any stand-out genuine photograph is going to lead the viewer to ask "yes, but is this real or AI?". I guess AI is going to mean there'll be a lot more stand-out "images" around in the future, but probably fewer stand-out "photographs."
Great news for pro work.
Will it still be on just desktop or will it be included on iPad
I’m hearing a lot of room echo. If you’re using a dynamic mic, it should not be so noticeable. Check your video editor settings to minimize it.
got any tricks on using gen fill when there's people in the picture?
I’d be interested in this too. Will say I haven’t tried much with people because it’s not so good. But the other day I was trying to take a person completely out of a photo. PS wouldn’t do it. But it did replace with a less obtrusive person I could live with. And actually you can’t tell in less you pixel peep.
Are there upscale algorithms to use on the lo-res results?
Sure, I use Topaz Photo AI to upscale, but it doesn’t help much with Firefly’s 1024px resolution cap.
How do I load older PSDs into the new version?
Do you have to be subscribed to the stock images for this to work?
Nope!
I heard that goes together with a new price plan for the use of AI features very soon. It won't be an extra , free feature.
Yep, a new credit based system for using the generative ai functions on top of the existing subscription payment.
Best PS Additions by far EVERRRRR!! 🎉🎉🎉
Now they have a credit system...that sucks
Why? The computational resources to perform these generative tasks is huge, especially when multiplied by the entire potential Creative Cloud subscriber base. Try using Midjourney. You can’t… not without paying for a plan that also limits you in some way.
Even though it will probably not happen, I truly hope one day when everything has been well-integrated, tied up and validated to their servers, Adobe will announce that they will charge users each time they launch Photoshop. Perhaps only then some of them will realize the dangerous direction we are heading on.
Let's say you have a complete photo set, how do you use it so that every photo looks like it's taken so that they "match" the style, light, color, and whatever else that keeps it looking like it was all done in the same shoot? Like who would want 50 photos with 50 different looking styles, light and whatever else.
Okay asking for pie in the sky i know but if Adobe AI understands that that is a building based on it's photo library and image recognition software, wouldn't it be great if it knew what building it was and, if it had other image references for that building, build and intelligent expansion with the environment that already exists at that location. I thought about this when i was playing around with a Seattle skyline photo that i have. the expansion was good but anybody who lives in Seattle would see that it was off a bit. I'm sure Adobe has city skylines in its library. Could they not then expand the skyline with buildings that exist within the skyline? Wouldn't' that be nice? Put that bug in your buddy's ear.😉
We are never satisfied are we? Lol
Eventually we won't have to leave the house!
@@joelwexler Yes take our vacation photos from the privacy of your home office. I had a wonderful time BTW but my wife is wondering who that blonde in all my pictures is.
Let's not forget that the new updated AI features in PShop 25, even though they are now approved for commercial use, are Only, 1024 resolution! That's a major drawback for anyone shooting Raw on High-resolution cameras.
Very true - although Adobe has stated that they will be increasing that resolution soon (but did not give a firm date).
I read somewhere that after Christmas Adobe will be selling credits to use the generative fill, and for each time you press the generate button you’ll use some credit. I was very disappointed to here that. Have you heard any rumours?
I did read about the new credits system, and Adobe published more detailed info about it here:
helpx.adobe.com/firefly/using/generative-credits.html
It looks like credits go into effect for paying subscribers on Nov 1, 2023, and you can check your credit balance in the Creative Cloud app (after Nov 1). Also, it's important to note that credits don't carry over from month to month, so it's a "use or lose" situation.
Still, this does not surprise me. Remember: performing these generative AI tasks requires a lot of computational power, and it can easily be abused without any sort of credit system. Midjourney has been doing this for a while now, and they don't even offer a free trial option anymore.
@@brianmatiash thanks that’s very helpful. It’s also not as bad as I thought. Photography plan gets 250 a month unless you’re subscribing after November.
Yeah, it’s pretty reasonable, especially because I don’t necessarily see photographers in particular blowing through their Gen AI credits each month. Graphic designers and digital artists? Possibly. But, I suspect that this tech has limited, albeit tremendously powerful, applications for photographers.
I have the complete cloud so I will get 1000 a month. More then enough. And CF will not cease to function if you run out, it will just become slower...how much slower is anyone's guess.
@@brianmatiash What a rip-off! I'm definitely only ever using it to remove things the eraser tool can't fix. Imagine the number of AI masks you could create in Lr worked that way - I'd immediately ditch LrC, forever.
Glad i watch this, so now i can uninstall Bate and just use the regular Ps. Thanks Brian
I have had trouble using the remove tool (I get a black selection nearly every time). Do you know if this bug been fixed?
Hmmm, I've only focused on the Generative AI tools so far, but will take a look at the remove tool in a bit. Although, I've never run into that issue before. The most basic first steps I'd recommend would be to uninstall and reinstall Photoshop. If it keeps happening, then Adobe Customer Support should be your next step. Let me know if the update fixes the issue for you.
Just tried it and it's working "OK" for me. I did notice that PS will fill in the whole area if say you are highlighting a bits of a fence to remove. Haven't gone any deeper on that.
Hey Brian. When can we start calling it "GenFill" and "GenEX" because ya know it's coming. Just remember I coined the phrase ..LOL
Ha! I’m all for it!
Awesome! I enjoy your channel...great job!@@brianmatiash
@BlueNoteJunkie5150 thank you so much, I really appreciate that!
only available to those with credit card and subscription for 30 per month for where I am, But its pay to use.
... "Generative fill roulette." lol Perfect characterization, but oh my goodness, what fun it is playing this type of roulette.
I am finding that the images it produces even with minimalist text prompts are amazingly accurate to what I envisioned.
I wouldn't be surprised to find that eventually the AI, if we have a subscription, will begin to learn what style of images we tend to like and give us ones akin to our preceding likes and dislikes.
Seems like a logical progression to me.
From a subscription loyalty, financial standpoint it would make me even more wedded to them and their tech.
Erik I have been using it as an alternative to hunting through stock images. It gets so close to what you see in your head if you describe it well enough, and it is way faster than searching stock libraries. The 1024 x 1024 resolution limit is a problem, but it's totally useful for finding (generating) smaller image elements. What makes it so doable is how amazingly good Photoshop's selection tools are now.
Does the gen fill now matches your actual resolution?
Brian talked about that many times in the video?
It's 2023, brah. Don't be that guy.
You no longer have access due to a violation of our terms
Ah how am I happy that I left adobe
Now their AI is out and they increase the price
Why even bother when there are better AI tools …
Fake world