Photoshop s NEW Offline AI No Credits No Internet

Photoshop now has a form of generative AI that is offline, does not use internet, no credits and this is big. You will find it inside of the Remove tool. At the top, if you click on Generative AI option, if you turn it on, there is a new device option. Before we even get into it, let's talk about how to activate it. First of all, you need to make sure that you're in Photoshop version 27.7 or above. And if you're not, you can easily update your Photoshop using your Creative Cloud desktop app. Just go to Apps. Here, you'll see all your updates. Click on View Updates. And if Photoshop has an update, you'll see it right there. Even if your Photoshop is updated and you select the Remove tool, this feature might still be grayed out for you. If it is, you need to make sure that you're meeting all the requirements. For right now, the Remove tool device mode which uses on-device generative AI has these requirements for Windows. If you're using Nvidia GPU, it needs to be 30 series or later 14 GB VRAM. It sets a high bar for requirements. Similarly for macOS, it may not even work with Intel based Macs and you need at least M1 Pro, not the regular M1. My MacBook has 16 GB of RAM. This requires 24 at least. Also for me, it was grayed out because my operating system was not updated. So make sure you update it to this one, version 26.4 or later. Even if you do meet all of those requirements and it's still grayed out, just go to Photoshop, Settings on a Mac, Edit, Preferences on Windows. It would be somewhere there. Inside of that, go to Image Processing. Just make sure Remove Tool Processing is set to Faster. And then inside of Performance, turn on Use Graphics Processor. Hit OK. Restart Photoshop to absolutely make sure those settings are in place. And then when you apply it for the first time, it will prompt you to download the model, which should be somewhere around 4 to 5 GB. Once it's done, you're good to go. Let's see it in action. Back in the magical world of Photoshop, and you can of course download these photos to follow along. Simply select the Remove tool. At the top again, Generative AI, turn it on and make sure Device is selected. Always create a new layer, keep it non-destructive and just loop around the object that you want to remove. I'm not speeding this part of the video. And this has a lot of perspective that it needs to fill. And oh my gosh, this was so darn fast. Now let us compare it with the non-device Remove tool. If we were to turn off Generative AI and not use any internet, this is the previous version. Let's see how this one does. Since this is not Generative AI, it would look around the image for it and have a look. It couldn't figure out how to fill those areas and what perspectives to consider. It is not generating anything, just looking around and copying and filling those areas with those. We'll get into how to make it better later. But this is an improvement from this and a big one. Now sometimes it might have issues with resolution and quality. Let's create a new layer here. And this time again, turn on Generative AI, Device. If we were to loop around the subject, especially right here on the ground, this is okay. But this might not be of the highest quality before, after. So in this case, what if we continue painting smaller areas again and again? Can we get this to look better? And have a look. We are getting clearer and clearer image. If any area doesn't look nice to you, just paint over that area one more time and it gets you something that looks plausible. Similarly here, the quality is not that nice. So I'm going to paint that area again, and that gets better. And keep in mind, all of this is happening offline without any internet, no credits, none of that. It's pretty cool. Now since this is Generative AI and not Content Aware Fill, every time you paint, it's going to generate something unique. For example, you want to remove this watch, even though this is cool. Let's loop around the watch, and it takes it away. And actually, it fills that with the sleeve. If you don't like that area, just do it one more time, and it will generate something entirely different, unlike Content Aware Fill. See, a different sleeve style. I'm going to do it one more time, a bigger area this time. And again, it generates something unique. And you can stop at what you like. And that is the advantage you get with this tool, a Generative AI engine that works offline. Now since this is also fast, depending upon the speed of your machine, you can paint multiple times. For example, in this case, if it did not get the stairs right, which it actually did, you can paint there one more time, and it fixed that. Now let's do a fun comparison between cloud and device in terms of quality and speed. Let's create a new layer. Right now device is selected and loop around the subject. Whenever you're removing something, do not forget their shadow as well. I'm not speeding this part of the video again. Let's see how much time it takes. And that was pretty fast. And of course, there are certain areas that it didn't get right, but we can fix that by painting again. But to keep the comparison fair, let's leave it at that. This layer was Gen AI on device. Let's turn off this layer and create a new layer again. And this time, change it to cloud. Do the same thing. Let's loop around the subject again. Again, I'm not making this part of the video faster. Now keep in mind, my internet is relatively fast, so the internet shouldn't be an issue here. But let's see how much time it takes. And talking to fill up this area is usually harder than actually editing that out. And oh my gosh, thankfully, it didn't take that long. Let us compare the quality. This is Gen AI on cloud. So Gen AI on device. As you can see, things are broken. It's not that high of a quality, but it's kind of good. But this is Gen AI on cloud. It created that stand, did this area so much better. And overall, I would say there is a big difference, even though it's slightly slower. And again, device. See the quality right there? Cloud. So much more better actually. But keep in mind, even in cloud, it doesn't cost you credits at the moment. But maybe Adobe is doing this because they want to charge you credits. And that's why maybe they released the device version. Anyway, that's just a theory. For fun, let's create a new layer again. And this time, let's use the best model irrespective of the credits. So I'm going to this time choose the Lasso tool and simply make a selection around the subject. Let's add a little bit of the shadow. Now we're not going to apply Nano Banana Flux or all of those models, but right click outside the canvas, click on Show Contextual Taskbar if you don't see it already, and then click on Generate a Fill and use Adobe's own models. Click on the drop down, Firefly Fill and Expand and click on Generate. Now this is supposed to be the best model there is for filling. And also, if you are on the pro plan, this does not consume credits for now. And there you go, we do have three options. One, two and three. Let's go with the third one. And it's actually pretty good. Even if you zoom in, we did not need that layer. So this was Gen AI on cloud. And this was Generate a Fill. Actually, even better. Here's three of them side by side, so you can pause and analyze them yourself. And this is zoomed in. Just look at the details, and you'll see all the difference. Now let's talk about some issues. And these might be just some glitch. If we were to select the Remove tool, and right now if we select Device, this should work offline, right? But if we were to turn off my Wi Fi and try that, it gives you a warning saying please check your internet connection. Now again, if we were to turn off Generative AI, this is back to work and it is working. But as soon as we turn on Generative AI, even if you're using Device, it doesn't let you work on your photo. But as soon as you turn on Wi Fi, this is back to work. Now it has connected. And now when you apply it, it should work and it does. Bingo. And again, it misses some areas up. But we can paint those areas again. Hopefully, it should fix it. Now as for another con, it sets the bar extremely high for the on device feature. For Nvidia, 14 GB of VRAM, Intel discrete GPU, 12 GB of VRAM separately for Windows. For Mac, even my smoothly functioning laptop will just not work because it requires 24 GB of RAM or more. The next con is a feature issue that it does not work with selections. If we were to select the Lasso tool here, and let's say we wanted to remove all of these wires from there. And in the contextual taskbar, if we show it, click on Remove, this does not use the on device Generative AI. It's using the cloud. It defaults to that. If you want to use on device, you would have to select the Remove tool choose on device from here, and then create a new layer and actually manually paint it because selections will not work if you need on device processing. Now this feature is definitely a step in the right direction. Even though it is not full Generative AI, it's the Remove function, it does generate details, it does create things and for one of the first times a major company has given you offline access to their Generative AI model that works in Windows and Mac and does not require complex installation of code or software. I hope that this model becomes better. It's only 5 GB, but we have had games of 100 GB. So imagine what it can do if it's that size. Also, I wish they create a full Generative fill version of this that works locally as well. Thanks so much for watching this video. I hope you enjoyed it. And if you did, make sure to give us a like and also don't forget to subscribe and not just subscribe, share it with your amazing creative friends and ring the bell so that you don't miss any other future tips, tricks or tutorials. I'll see you in the next one till then stay tuned and make sure that you keep creating.