What's New In Photoshop Beta (April 2024) - MAJOR AI Improvements

🔑 Get my UPDATED Lightroom course: www.matiash.com/lightroomever...
Adobe recently updated Photoshop Beta with major improvements to its Generative AI technology powered by Firefly, and I'm going to focus on why photographers should pay close attention to it.
Chapters:
00:00 - What’s New In Photoshop Beta
00:47 - How to install Photoshop Beta
01:46 - Improvements to generating an image using Firefly
04:52 - Major improvements to Generative AI quality with Firefly 3
06:44 - Using Reference Images
08:45 - Improve quality of generative AI using Enhance Details
09:37 - Generating similar options
10:17 - Using Enhance Detail with Generative Expand
12:17 - Improvements to removing and replacing the background
#generativeai #adobefirefly #photoshop

[@fabienb]

There is actually a major “bug” in the new generative fill, as it seems to now be entirely intended to show something: I used to use it to clean images in the cases where content-aware fill would fail, and it used to do a pretty good job at removing people or extending surfaces. But now it hallucinates all the time, and rather decides to create something extra when not needed, which is extremely annoying. I removed people from in front of a brick wall, and while this used to be a simple task, the new version doesn’t allow me: with every try (blank prompt) it created a graffiti, a statue, other people, a steam engine… When I prompted ‘remove people’, it even added a bathroom sink… I hope this gets fixed soon

[@roderashe]

Try the remove tool too

[@jaspergoodall3206]

That’s very worrying. The only time I ever use generative fill is with a blank text box. Saves me a lot of time when needing to retouch a distraction from a busy area, say, a tree stump or log in a woodland scene. If it starts adding stuff it becomes useless! I wonder if it makes a difference to do the fill incrementally? I do fills in bits because my files are quite high resolution so too big an area looks obviously blurry. Smaller areas retain better detail. Maybe this method could make a difference? Nevertheless, Adobe ought to set a default ‘no hallucinations’ for those of us who leave the box blank.

[@fabienb]

@@roderashe that's another tool I use a lot, but there are cases in which I found genAI to work better, particularly if I need to recreate something that is not otherwise visible (i.e the bottom of a column)

[@fabienb]

@@jaspergoodall3206 hopefully it is only temporary. I do the same as you: since the resolution is not satisfactory I do my fills in batches (usually 500x500px) and before this update I never had an issue, sometimes even with mindblowing results. One terrible result with this update was on a simple fill of a marine scene, where it only really needed to fill with water (and content aware was way too soft): genAI started adding all sorts of unnecessary content to fill the space.

[@tdalton854]

Once again, a nice calm way to explain new stuff. I just added the beat after watching. I was most interested in the generative expand working here. Beta or not it almost always amazes me. And that enhance button worked in a good way. I don't use this often but I like how the start from scratch works. I didn't compare to ps 2024 but holy heck just choosing photo is very noticeably better. Thanks

[@valleyfella]

Thanks Brian for another really helpful video. I watched it shortly after doing a focus stack between LR Cloud and Photoshop. It’s a really clunky process and as much as the development in AI is interesting and useful, it’s frustrating that Adobe seem to be ignoring the basics like opening multiple photos as layers from LR cloud. Hopefully they can give some assurance on this soon. Thanks again for the video though. You are a really clear and helpful teacher.

[@polystormstudio]

I like how you explained all this better than the 2 other videos I saw about FF3 and Beta's new features.

[@brianmatiash]

Thank you!

[@DuaneCaldwell]

Very nice demos of the new improvements! Thanks for sharing them, Brian!

[@brianmatiash]

Happy to help, Duane!

[@DuaneCaldwell]

@@brianmatiash, I've been appreciating your helpful advice and tips since The Lightroom community back in the old G+ days! Been following you ever since.

[@jaiharji4840]

Just discovered your content. 🎉🎉🎉🎉🎉 Something In sure many of us could do with a video on is how to use photoshop to create product photography for social media etc… using existing product/pack shots

[@ralphpayne6259]

Brilliantly clear video and Firefly 3 access is, indeed, BIG bonus! Like others, can't wait until Lightroom gets similar quality improvements! Thank you, Brian!

[@mbhullar]

Excellent tutorial, hope to watch more of your videos.

[@thewebstylist]

All Amazing additions! 🎉

[@erik1836]

Thank you very much Brian. This was very helpful to me.

[@brianmatiash]

My pleasure, Erik. Glad it was helpful.

[@alanplatt888]

Fantastically enlightening as usual, thank you

[@brianmatiash]

Thank you as always, my friend!

[@DanHughes]

Smooth and easy listening with Brian Matiash. Pithy and informative; thanks for making this stuff.

[@brianmatiash]

Rock on, my man. Always a pleasure to see your name pop up. Miss that face, too!

[@goldspike1]

You do an excellent job of explaining the technology the first time around so I don't say "wait a minute what did he say" and I don't have to stop and run it back 4 or 5 times. Thanks

[@brianmatiash]

I really appreciate that! Seriously, it’s one the nicest compliments you could pay to an educator on this platform.

[@user-kw4xw2fc4o]

First time on this channel. Very nice video. Did the latest beta improve the ability to clean up old photos and colorize them? I’m pretty new to PS but have been less than satisfied with some of the quality of the restoration tools.

[@NeilFarrimond]

Great overview! Looks like some amazing updates.

[@brianmatiash]

Thanks, Neil! I’ve got your ecamm Zoom video to go through later today.

[@NeilFarrimond]

@@brianmatiash if you want to do some testing over the next few days let me know :-)

[@brianmatiash]

@NeilFarrimond count on it.

[@jackprice7828]

I am loving the new PS tools. The one I find somewhat frustrating is select subject or remove background. Depending on the subject, but especially with people, it tends to not detect well when there are complex pixels with light and dark contrast areas.

[@sima1203]

Excellent, thank you very much from the bottom of my heart

[@TC_Conner]

Very good and informative video Brian. I installed PS beta once and it made itself the default image editor whenever I clicked on an image. I deleted it and haven’t tried it since.

[@joelwexler]

I had same annoying thing.

[@royceahr]

Brian, was wondering if you encountered an issue with the Beta not initializing if you have plug-ins installed? I installed this new version over the top of a previous Beta but found that it would not initialize so I tried the Shift click method to start which disables all Plug-ins and it found that it initialized and works as you show but without any plug-ins.

[@jayeshtopiwala3219]

Any update coming to lightroom classic ?Denoise Ai was introduce in May 2023.

[@BubbleGendut]

Rather than fill in one go use selections to generate several segments it better matches the resolution of the original image

[@fashopimpin]

I love generative fill. You can lasso hair with fly aways and it’ll lay them down. Just lasso click generative fill then generate don’t type anything. I even removed an entire forehead of acne with generative fill. Not typing is like content aware with a lasso

[@brianmatiash]

Oh man, I agree 100%! There is no contest in terms of the results between removing an object/distraction using Generative Fill vs a remove brush. Major part of my editing workflow.

[@fashopimpin]

@@brianmatiash Amen!!!!

[@fashopimpin]

@@brianmatiash When they added it to the iPad app I was here we gooooooo!!!! Lol

[@michaelmcdonald3345]

Good video. I tried the last PS Beta. However, it did not communicate well with my LrC program. ??? I ended up deleting and reinstalling both to get it to work right. I like the improvements, but I think I'll wait for the final version.

[@aviatorman80]

How does all this help with upscaling a 3 MP photo to 12 or more MP?

[@MOWFWDU]

I agree expand is better. The thing about the example you showed us in generate background is that the select object is easy. In the real world it's not and as far as I am concerned I wish adobe would bring AI to the selection tool.

[@jimharris3889]

It's still got a ways to go to be "good" I had an ocean landscape scene and I wanted to add a "pirate ship" to the scene just for fun, the pirate ship looked like a cartoon image

[@TomatenDK]

I would looooooove if we could a Generative AI version for Lightroom.... pleeease please please Adobe

[@brianmatiash]

I suspect that we’ll see something like that eventually.

[@TomatenDK]

@@brianmatiash eventually.....we have waited too long already imo. But fingers crossed 🤞

[@RandyPollock]

I would much prefer focus stacking in LR than AI, but I guess with AI I don't need to focus stack just tell the overlords what image i need :-)

[@frand6822]

Thank you for another very informative and helpful video. As for the topic, I've made the decision to not use generative AI. I'm a photographer and though I no longer shoot film, I'll stay old school as much as possible. Yeah, I know, AI is assisting me in LR and PS in many ways, but this is way beyond where I want to go. And, I disagree with "saving photos" with the expanse "intelligence." Just go out and shoot another photograph. Oh, did I mention that I'm a photographer? 🙂

[@graphguy]

The only improvement I want to see if to drop the draconian subscription ONLY model.

[@jasonkwilliamson]

The day that AI can seamlessly install an existing image into a photograph and work around it is exciting

[@tomberault4434]

but now i can't have the generative fill interact with MY IMAGE by going the quick mask, filling with a percent of gray, leaving quick, put in my prompt and watch the generative fill INTERACT with MY IMAGE. I SHURE HOPE ADOBE DOES NOT TAKE AWAY THIS FEATURE as they have with the Beta version

[@Woici]

Unfortunately the last two BETA updates didn't work for me, I had to go back to version 25.7.0

[@brianmatiash]

Hmmm… I’d reach out to Adobe Support to see what’s going on. It should work fine assuming that you have compatible hardware.

[@Woici]

@@brianmatiash I had a long chat with Adobe and was repeatedly told that support for the BETA was limited. There was no choice but to go back to the penultimate version. My hardware is top of the line... Regardless, I'll try to activate the BETA again with the next patch 🙂

[@GoodaySunshine]

Generative fills are taking longer and longer to load each day, even though I have plenty of credits and always update the software 😢

[@JamesHarris-]

I'm waiting, hoping Adobe will stop trying to censure our private images. The warning, "This doesn't meet community standards...." is exasperating, especially when it seems to be due to showing a female elbow. I've had this happen repeatedly to the most innocent images. And this isn't any copyrighted material but private family photos. There are work-arounds such as masking off the offending elbow, then running AI on the 'Taliban-Approved' remainder of the image. I am an adult and I do not need Adobe trying to control how I edit my private family photos. How about this: Instead of treating their customers like children, why not put some of their resources into improving the way filters clean up and align pixels in old photos being restored? You know something actually useful to the customer. Consider the following: If Adobe is censoring customer's photos, I wonder if they are retaining and storing them without permission? They have already been hacked and customer data stolen; what will they say if our images are also stolen?

[@alangibson3293]

All this AI crap you people cream your genes about is not photography. If you take a picture in your over expensive camera that is not what envisaged, try a different angle. AI is not photography, so don't call yourself a photographer. Maybe call yourself a compositor instead.

[@josdenis3684]

A digital compositor perhaps?

[@williamsadler6467]

This could be fun, but don't call it photography. Once you add something to an image, it's no longer a photograph. So in a sense you can't improve your photographs, you only destroy them.

[@brianmatiash]

Does that also go for using a healing or remove brush? Have you ever used one of those tools to remove a distraction or even a dust spot? Is it a photograph then?

[@williamsadler6467]

@@brianmatiash Dust spots are ok, but removing objects? I would say that is not a true record of a scene. With the rise of AI I think it's important that we distinguish between what is real and what is not.

[@startupstoriesbyankit]

Does beta export the generated image in the canvas DPI size? My canvas is set to 300 DPI but the export is 120 DPI - @brianmatiash - Do you know a fix of it? The normal PS generate quality is really bad but exports in the right DPI. So that is why I want to use beta (For quality) but it does not export the desired DPI - any thoughts? thanks