Stable Diffusion - Zoom with Outpainting with Fooocus

[@aquietmonk]

I know this is an older video but I've just found your channel and started catching up. Thank you for making these great tutorials it's helped a lot. A top tip for outpainting and avoiding seems is to paint the edges you are expanding out from (for example the left edge or right edge) with a small mask brush and wiggle it slightly so you don't have a perfectly straight line. This helps the AI drawing denoiser blend the edges better and avoids the seems. Peace and love from the UK.

[@Sounds-Books]

I think you are one of those explain foocus really good and simply. Thank you. And i am subscribed. ❤

[@KLEEBZTECH]

Thanks for subbing! And thanks for the feedback.

[@Bamigaru]

new day, new fooocus class! great video and explaination as always. Thanks!

[@HayleyHadeez]

I should just sit and watch all of your videos. I spent ages trying to find an expander to modify an image in the wrong ratio not realizing outpainting would do it. I assumed it was used to paint out stuff you didn't want. Doh!

[@KLEEBZTECH]

I would recommend it and not just because I want more views. lol. Many of my videos I try to show techniques that can be used for many things and give a better understanding of how Fooocus works. For example the one I have on good text reveals some details that can be used for other things besides text.

[@HayleyHadeez]

@@KLEEBZTECH yeah I'll definitely be using it from now on. I'll revisit the image in question and see how it works. I was using a skinny ratio to get a long body shot but my other images were closer to a widescreen effect. I got it working with a free site but I'll definitely try it with Out paining. Thanks for the video!

[@Dalin_B]

Awesome man. I actually recently discovered the power behind the outpainting! I am however having issues with the seam in the picture. I've had to use Photoshop to take care of that

[@KLEEBZTECH]

I have not used outpainting that much so still experimenting to find ways of limiting the seam. Have done the same and used Gimp to fix them. But just like you I have had my eyes opened about outpainting and what can be done with it.

[@Dalin_B]

@@KLEEBZTECH Man. The outpainting is POWERFUL. The thing is, the Refiner switch and the respective field are crucial. Also, I've found huge difference in outpainting one-sided at a time vs multiple sides at once.

[@KLEEBZTECH]

Absolutely agree on one side at a time. ​@@Dalin_B

[@DanielSchweinert]

Great video series! Thank you! I recently started using Fooocus - coming from A1111 and Invoke AI. Would be great if you make a tutorial about the 2x Upscale in Fooocus mimicking Magnific AI. Im sure there are some sliders in the debug mode to get better results out of Fooocus Upscale.

[@KLEEBZTECH]

Yeah I did a video on upscaling but nothing in-depth. I plan on going back after more experimenting since I have traditionally not used the upscale in Fooocus. As for mimicking Magnific AI, I doubt it can come close. Javi has done a great job with Magnific AI. My issue is upscaling and recording on my system. But I have a different video card with a little more VRAM that I plan on trying out in a few days.

[@DanielSchweinert]

@@KLEEBZTECH Yes I saw that video and it's a good guide to get starting. I've done some amazing upscales inside of A1111 with Ultimate SD Upscale and some prompts but of course it's a PITA. Fooocus is soo much easier to use. Will also try to get same results by combining SDXL and SD1.5 models as refiner.

[@TheCopernicus1]

Subbed, awesome content mate!

[@KLEEBZTECH]

Much appreciated!

[@TheCopernicus1]

I have moved from A1111 and no regrets, the creative freedom for what you need this Fooocus is suffice for most applications! @@KLEEBZTECH

[@KLEEBZTECH]

I started with that and went to ComfyUI and then this.

[@historiadevelada]

Awesome tut Thanks!

[@KLEEBZTECH]

Glad you liked it!

[@puja1985]

Just like Midjourney but much better, you can upload your own image instead of relying on generated images to use the expand or zoom out tools, as in Midjourney , bravo fooocus,, and thank you for the video , would love to see in depth tutorial on the faceswapper feature here..

[@KLEEBZTECH]

Will get there and hopefully soon. I have also been testing the negative prompts and can say they have much less impact than I expected. When I was using other tools like ComfyUI the negative had more impact on the resulting image. Now it is almost impossible to see the difference.

[@Arkapratimbhatt]

Thank you

[@musicalstation101]

Very cool video can u do like foocuse and pika lab to make very perfect video please

[@KLEEBZTECH]

I probably will do some AI video stuff at some point but not sure how soon. I still feel the video ones need to get a little further along. But will add to my list of suggestions. Thanks.

[@musicalstation101]

@@KLEEBZTECH and when u gonna do the ai influencer and how to make it so real

[@a_vin7644]

OK, this is for zoom out, what about zoom in? Can I upscale only part of image as new image generation?

[@KLEEBZTECH]

You could but you would have to crop it with other software and bring back into Fooocus.

[@masterzed1]

Hello, I'm looking for an out painting that is similar to in Paint anything, meaning I want to create a mask within the drawing and then do the out painting. Can you provide an example using masks? I've tried doing this with in Paint anything, creating a mask over a person, and then modifying the background. However, it does it horribly, as it blends the original image as if it were part of the newly created image. I want to add a background with a robot behind a person, and when I input that description, what the AI does is only create a robot using the person as if it were part of a leg or a part of the robot.

[@KLEEBZTECH]

Would this be what you want? Or help? kzbin.info/www/bejne/nHK1lY2BZ8t2htE

[@ibrahim__alsayed]

Thanks

[@KLEEBZTECH]

Welcome

[@zeeshankazi-r8g]

how can you do reference generation? adetailer in foocus?in colab

[@KLEEBZTECH]

Do you mean using another image for a prompt? If so I have video here. As for Colab I know there are some limitations but not really familiar. kzbin.info/www/bejne/aJm6YYBjjpaFbM0

[@zeeshankazi-r8g]

@@KLEEBZTECH i saw your this video i learned alot thanks for making such videos i am working on image2image source image same face--- faceswap target image same clothes - i Dont know same pose - canny same background - i Dont know also, how to install extensions like afterdetailer. i dont have nvidia computer working in colab

[@CaseClosedStudio]

There is no zooming explained in this video, adding on to the sides is expanding the image, aka outpainting. Zoom is a completely different thing.

[@KLEEBZTECH]

Zoom can be in or out. Similar to how Midjourney uses the word in their interface which is what Fooocus is designed to be similar to.

[@CaseClosedStudio]

@@KLEEBZTECH You are incorrect. SD/Fooocus doesn't use the word 'zoom' at all in their documentation. Midjourney only uses that term for 'zooming' the entire image, not expanding a single side - that's the arrows in MJ (never referred to as zoom) and outpainting in SD/Fooocus. If you have a reference to the official documentation that uses the term 'zoom' in the way you're describing, please link it here. I have no problem being wrong but saying you're correct just because you say so is not valid. Zooming in SD/Fooocus would be using the top/down/left/right radial buttons all at the same time and that usually gives poor results.

[@wernerblahota6055]

👍👍👍👍