Good Lord, just after the 2 first minutes, you became a hero to me! I seriously didn't know that the contextual task bar was made to sit just under what it had created. And it was driving me crazy! Thank you!
Came for Generative AI ideas, stayed for all the general PS editing tips (e.g., at 28:30 using offset + GenAI to create seamless 360 panos) 🤯 Thank you!
I learned so much by watching your artistic process. I've realized that I need to train my eye to focus on the colors and saturation of shadows- I'm in awe at how naturally that comes to you.
Thanks for the great training video. Funny thing we have an Icecream store in Fort Bragg CA that actually makes mushroom Icecream. It’s made with candy-cap mushrooms and is really good. Has a maple/toffee flavor. Thanks again.
Thankyou for showing us the mistakes the program makes and how to fix it. Stuff like that gets me stumped for hours trying to figure it out on my own :)
Thank you for going over these ways to utilize the GF feature in PS beta. I think there is some promise to occasionally using the features in beta, but for example, the depth of field regarding the mountains could have been done many times faster just by selecting them and gently applying a blur filter of your choice, then deselecting them and softening their edges and possibly tweak other areas, if desired (or some combination thereof). 😊
Loved all the tips, especially how it wasn't just AI, you actually told us how to incorporate it with our current workflow. You're a great teacher. Subscribed :)
I had no idea how to do a lot of this in Photoshop until watching your video. Thanks for some great tips! I use many of these tricks in Lightroom though, thanks to the masking feature. One thing I would have done with the ice cream cone is to continue it through that void in the mushrooms, since that’s the angle the back of the cone is going in and you expect part of it to show up there.
There a better way to bring these koi fish under the water. Press Q to activate the quickmask Take your brush and paint with a greyish color to create a semitransparent selection. When you are satisfied with the selection, press Q again and use your prompt in generative fill. The Ai will generate the fish but less intensified, it will blend much more with the original image. There will be no transparency though, the generation is just less intensive.
Such a great video in generell. Great production quality, nice to see your face in between tips and just great tips for PS too. Please keep it up, outstanding work.
Excellent information. I have a problem with generative fill that others might be encountering and you might know a solution . This occurs when using GF to remove something. For example remove a car . Sometimes instead of entirely removing the thing I have selected GF replaces the thing with another version of it . For example instead of removing the car it replaces it with a different car . I use the technique of leaving the text box empty. Is there something whereby every time it will always remove the selection content instead of replacing it with a different version of same thing. Thank you
I've run into the same situation and after some testing, I've found that it usually replaces instead of removes when you've left a shadow or reflection outside of your selection. So, when you make your selection of an object to remove, be sure to select its shadow and reflection as well before running generative fill. That should help!
Great video. Loved to see the process, the creative thinking and your ideas to execute your imagination. I watched the video full 35 mins and can say I wasn't bored at any second of it. Great job. Very helpful content. Value creation. Thanks for this.
Haven't really been updating my photo editing techniques for ages and i learned so much in this video that wasn't even about AI. Your techniques are increadible 😇 The generations themselves are still really bad though. Tried adding "old rally car on the grass under tree", even gave it some context, and the stuff it came up with were just some random car parts scrambled together.
Love the use of AI as an editing tool rather than just making up fantasy stuff. What a great video. And the tip on avoiding violation triggers is gold, I actually had an outpaint that kept generating these when I was trying to get a little more parquet flooring beneath the main subject. But simply adding "floor" as the prompt did the trick! The only thing still missing for me is the AI helping out with matching a subject to a background in terms of tone, shadow, white balance, etc. I wonder if there is a way to convince the AI to help out with that.
These are great tips, thanks you. How are you getting around the resolution limit for your generative fills. You are using large images. What does the generative fill look like when zoomed in. Or are your selections small enough. Just wondering.
That was interesting thanks Rikard, some useful tips! I’d like to know if there is any way of adding a missing family member in to a group photo using Generative Fill convincingly?
using the tilt shift makes the background and foreground have an even blur which would be off from real life you want to fade the lower blur out more using other ideas
Yes!!!! Quick mask is a hidden gem in Photoshop. You can use it to adjust your selection in all sorts of ways that are impossible with any other selection tool.
This might help: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-cOk2YpKDv9k.html but at the end of the day, you're always going to get the best results doing it visually. I have a course for color correcting and grading in Photoshop that has a lot of projects all centered around color matching, correction and grading. www.nucly.com/photoshop-color-grading-masterclass-new
The one thing i don't understand is how you are getting realistic looking generative fills especially people. Anytime there is a face involved it results in a Pollock meets Picasso result?? But on youtube i see so many that generate great and realistic looking people and i would even use the same prompts. What else is it grabbing from that i could help this? I'm not really interested in people but say change of clothes, adding clothes, or other objects that actually have a realistic look?
Yes, generative fill analyzes the entire image every time you run the tool. Simply add your subjects into your file first and then run the tool on the background.
I do not understand how you make an adjustment to the quick mask - it does not load it, but adjusts the whole picture. I know I am doing something wrong as I managed to follow once by total accident, but do not know what I did differently!
Hit Q to get into quick mask mode. When in quick mask mode, you are manipulating a black and white channel, visually represented as a red overlay. When you hit command-M to bring up your curve, you can adjust this channel. However, as the curve adjustment pins your blacks and whites, adjusting the middle of the curve won't do anything. You have to adjust the black and white points of the curve-these are the points at either end of the curve. When you adjust these, you will see the value of the red channel overlay change. The more red it is, the less strong your selection will be. Let me know if that solves it for you.
I'm about to put one out that dives deeper into intensity and the quick mask. But essentially, the quick mask tools temporarily turns your selection (running ants) into a mask (black and white image) that you can manipulate just as you can any other black and white layer-with blurring, brushing, smudging, curves, etc. When you then go out of quick mask mode, the changes you've made to the black and white image (mask) are transferred to your selection.
@@nuclylearn ah I see... I mean.. basically its like a layer mask except it exists in channels rather than as a black and white mask in the layers. Considering the power of using generative fill on a layer where it looks at whats inside of a selection.. they should make it a selection where you can tell PS to take whats inside the selection into account, or not etc. The fact that it ignores whats inside a selection can also be a large weakness. I kind of think PS needs a retooling period. For instance, why is it that most of the retouching/effect/ work I do on a photo is easier in After Effects? PS has almost no real plugin/tool library. While literally every single AE effects tool should be available in PS for one frame. I actually remember you used to be able to take PS plugins and force AE to use them. But now it should go backwards. Remember when photoshop had a paste warp? You could paste into mask and it would warp what you copied to fill in the mask? Circa 1996ish
@@jonahoskow7476 I agree that PS needs some retooling. A lot of tools and features have been added on top instead of properly integrated. It's a bit silly that we have to go into a different workspace for "Select and Mask" when the only additional tool the workspace now provides is the refine edge brush. Why not just add the refine edge brush into Photoshop and get rid of the Select and Mask workspace? Filters is also a good example-some are in the filter gallery, others in the neural filters, blur filters have their own workspace for no good reason, etc. The problem is that if they do a retool of the software, a lot of people are going to be up in arms. But ultimately, it's probably time they did.
Have you faced the wrong sizes of the generated background? Im doing product photos for website. I have an object that is around half meter diameter, its a roll of a garden hose. I remove background and ask photoshop to put it on standing on grass. And photoshop creates close-up background and eventually it looks like a macro shot. And my half meter diameter roll of a garden hose looks like 10 centimeters. Any advice to specify size of generated background?
Unfortunately I don't know of a way to dictate the size of objects consistently. The best way I've found to change the results of the AI is to change the context in which it operates. So if it is generating at the wrong scale, try to add more elements of the correct scale into your scene before generating. You can then remove those objects after you've generated.
@nuclylearn Thanks for the advice, I have found that if you make the selection with your original background turned on, eventually you will receive more realistic results as AI reads the context, as you said, it really helped me to place my products on white background. It realistically creates shadows, however it doesn't work with placing objects on the grass. I mean sometimes it does work, but 80% of generated backgrounds are trash
Even I know I will get scolded, this video is for photoshop pros. I liked it, but as a beginner there's too much curves, layers, twi images combined, blending, etc.. Nevertheless I subscribed cos I know I can learn a lot
No, it can't. It's attempt at writing are only usable in small thumbnails. You're better off just finding a good handwriting font and typing whatever you want.
I am seriously worried. i have a $2000 GFX lens that can do this, but why should i risk taking that with me, if i now can do this with a phone + photoshop beta????
