► UPDATE (as of November 3, 2023): Generative Credits are still unlimited. Here's what Adobe's website says, "For a limited time, Creative Cloud, Adobe Firefly, Adobe Express, and Adobe Stock paid subscribers won't be subject to generative credit limits. To learn when credit limits will apply, check back here after January 1, 2024." Also, this feature in the plug-in is 100% free.
You are my teacher Umesh I have learned a lot from you. I know nothing about Photoshop a few years back. With your videos my journey started and it's still going. I am an e-commerce reseller I don't pay any third party company for editing instead I do it on my own. I learned from you. Huge respect and gratitude for you. 🙏🏼
You are the only channel I watch for Photoshop content. Your content is delivered so well, and easy to follow. I am excited at all the new things that Adobe is coming out with, and you break down these new features amazingly. Thank you for teaching us!
This is the BEST Photoshop channel in RU-vid by FAR. Thanks a lot for everything you do. I wish I could pay for that amazing plugin to automate some steps of compositing, but I believe I have them pretty well learned thanks to your previous tutorials. Thanks again and I hope you're well :)
Thank you Unmesh for this addition to your Compositing Tool Plugin. The complete tool is so worth the cost for anybody serious about Photoshop compositing, it makes it so easy.
I've always admired the way you approach Photoshop techniques. This tutorial on generative fill was insightful and well-delivered. Thanks for consistently sharing your knowledge with us.
Impressive! AI potential pushed even furter 🤯 I've been listening to your content for more than a year and I must say... You deliver man! Big thanks for everything you are bringing to the Photoshop community 🙏🏻
I downloaded the tool and have tried it on over a dozen different images and it mostly works. The generated sections are often still a little fuzzier than the original image, although much better than if you just selected a large section of your image to gen fill. You will almost always see distinct lines between your original image and the gen filled images when you're looking at full resolution. If you're shrinking the image down for the web this probably won't be an issue. I usually go over the lines and do a Content Aware Fill to blend them together. The other problem is that if one of the generated sections looks off, you can't easily go to that one and choose one of the different three options since every generated section after that was based on the first of the three options for each section. You'll have to just redo that section overtop of the one generated by the tool. It's good, but not perfect. If you're willing to deal with the minor annoyances I mentioned, it can still be helpful.
Man, your sync with those people remarkable! I got their email and your video in my YT feed like at the same time! As impressive as the tool itself! :-)
This method is very similar to how I approach editing, mixing and cleaning up my recordings for my songs, if I need to remove noise in the background of an area, I take a sample of a large portion of the sound I need to remove rather than a tiny bit to use as an example for it to replicate the inverted sound to remove it. This makes a far superior result - much like what you did here. Also, it is like when I record my songs at 24bit 192khz, this means when I turn up the volume on quiet bits, there is no noise but if I recorded at a lesser bitrate it will be hissier. Same thing here when you used the entire image as a sample size vs using a smaller bit.
This is great!!! I did see that if you do it by spots, it doesnt' carry the effect to the entire image (Of course), like the shadow of the girl does show up when done at once, but not by pieces :)
OMG ! I didn't even realise there was a Pi plug in!!!! Absolutely awesome. You legend. Thanks so much for this tutorial as well. So is it best to extend the canvas for gen fill rather than crop and fill ? Thanks again Unmesh 👏🏽👏🏽👏🏽👏🏽🙏🏽
Genius! In your example with the woman, the process you created used maybe 20 "boxes". Does this man that (after November 1) 20 credits of Generative Fill will be used? Thanks again.
You are a sweet baby angel for this!!!!! Seriously I saw that smaller boxes did better but the time it took? Yikes! Another great share for the community. Also, I didn't realize the Nov. 1 for credits. I've already been kinda careful.
Hey Unmesh. I tried your tool yesterday and realized that it doesn’t work for generating *inside* an image. Is this going to be added as well? Ps: love how big this channel got and how professional looking your videos now are. Your photoshop skill has always been admirable
Just downloaded and tried it out. Worked great. A few lines that had to be cleaned up but that just took a few seconds. Interesting that I had a few lines visible in the vertical but none in the horizontal.
I'm wondering if there's a workaround to prevent the generations from getting messed up. Would it perhaps be possible to generate the whole thing in one go first to get an overall idea of what the picture is going to look like and then use your technique to sharpen those blurry areas? 🤔 but either way, great job, Unmesh! U rockin'!
@@Pictureinstruments-PI Hm yeah makes sense. The main issue with Firefly is that it doesn't understand your prompts. It interprets all inputs as requests for generating new content, so you can't directly make it sharpen things. If it were possible tho, it might still be overridden by the logic of the AI. For example, if you have a blurry area and want the AI to unblur a specific part of it, Firefly might try to create a smooth transition, which looks more natural since such harsh transitions between sharp and unsharp areas aren't natural without a change in the depth of field. Alternatively, it might refuse to sharpen anything at all if it doesn't see a logical reason to do so as is at the moment. It will just generate some incohesive stuff if you type in "sharpen" or "unblur". Funnily enough, it will even tell you "unblur" isn't a commonly recognized language and not do anything altogether lmao. But then again, if it could understand prompts like "sharpen", there'd be a potential workaround. You could generate the entire image first, then select the first 1000x1000px square, temporarily remove the rest of the generation, and merge that selected square with the original image. Then, you could let the AI sharpen that specific part. Repeat this process for all other areas. This way, you'd trick Firefly's logic component because there are no nearby distractions that could prevent it from sharpening as you want. Hopefully, Adobe will improve this in the future or just give us higher-resolution generations right away lol. I hope I made sense XD
Thanks for this awesome plugin. I think it would be better if we can rescale or resize 1 part smaller before we use the Hi Res Gen Fill plugin. This will maintain the original size but we expand the image wider.
Thanks so much for all your video's which helped me to learn a lot about photoshop during the past months. I update my photoshop with generative fill hoping that will help me as a professional. Unfortunately, the result I got from generative fill specially when I try to change the outfit of some models are exceptional disappointing and surprisingly I got much much better result from photpea. I believe it will be great if you can post video regarding how to change outfit using generative fill. Thanks
An option I'd like to see is to generate the entire area first, chose which of the 3 you like, THEN generate over that for more detail. That way you're more likely to get the result you want with less credit usage!
