Thank you SO MUCH for the explanation of what is actually happening when u click or button or move a slider. I still dont really understand how stop at affects and image though. I get that it stops influencing at x percent of img gen but thats not very helpful as that doesnt tell me how its gna affect the img. What is the provess of how the img is gen’d? What gens first and what gens later so i can think concretely about what happens each percent step of the generation? What details are added in the first 10 percent vs the last 10%? For ex
Hey Joanna, thank you for the great question! The "Stop At" slider controls how far the AI goes in refining the image. Early steps (like the first 10%) focus on big shapes and overall composition. Midway steps add details like textures and shading, while the last 10% sharpens fine details and polishes the image. Stopping early means you'll get a less refined, more abstract result, while letting it go further creates a more detailed image. Hope this clears it up! This article goes more in-depth into the entire process: www.altexsoft.com/blog/ai-image-generation/
Have a look at the discussion of the Github link I shared. Here is an example comment: github.com/lllyasviel/Fooocus/discussions/557#discussioncomment-7226225 If you scroll further up & down you will see other people having similar questions and just see what they tried out. In your case, if you want to have a dog in watercolor style, use 1 image of a dog and 1 image of a dog in watercolor style. Just search on Google for an image you want your end result to look like, save it to your computer and just use that as input. Then, play around with the stop at & weight of both images. Start with both settings as you can see the person used in the link I mentioned above. So, PyraCanny for your dog image, with stop-at at 0.4 & weight at 1. And ImagePrompt for your dog in waterstyle color with the same stop-at & weight as the other image. Regarding your styles in the panel on the right, just keep FooocusV2 checked. Let me know if that helps?
Looking for a fresh take on image creation? Explore Image to Image in Focus and discover how AI can help you generate unique and visually appealing images. #AI #ImageEditing #Innovation #SmythOS
There are actually quite a few different resolutions/aspect-ratios you can choose from. For this, just tick the Advanced option below the text box for the prompt and go to the Settings tab. There you can find for example: 1536x640 1664x576 1728x576 If you want to have an even bigger resolution, you can just drag and drop your desired input image into the upscale box of Fooocus and then just upscale the image. If you don’t know how the upscaling works, we have a simple video about that on our UA-cam channel. Let me know if that helps.
There is no "best" checkpoint per se. It all really depends on what you want to achieve. The default settings of Fooocus already give some great results to achieve human realistic images. Otherwise, you can try using JuggernautX as base model or instead of opening up Fooocuse normally with run.bat, open run_realistic.bat Let us know if that helps :)
@OfficialGTDuo i try it but ondy portrait for face is good quality, when make full body image then always blurry, i try both realistic and run.bat, even i try upscaling, image size go to 25mb but image still blurry, skin texture like hand painting
Have you tried Fooocus inpaint in that case? We have a tutorial about how to use Fooocus for improving details in an image like the face: ua-cam.com/video/P4N1HRWj2Jo/v-deo.html Also, you can have a look at this Github discussion where there are some more examples on how to improve faces (scroll down to the section "Advanced: Detail Fix (face, hand, eyes, ...)"): github.com/lllyasviel/Fooocus/discussions/414 You will be astonished by the results you get when using the inpaint function. Let us know if that helps.
With such a cloud service it is possible that you have too little storage or GPU power on lower price tiers and might have to upgrade to a higher pricing tier to get it to work properly
In the folder you downloaded for Fooocus there should be a file called run.bat which you just have to open. It will open up a command prompt window, which does some setup and then it will automatically open up your default browser where you will see the UI of Fooocus. Alternatively just check out our video which shows how to install and start Fooocus here: ua-cam.com/video/s10LL9cUQmQ/v-deo.html
LOL but what a beautiful work it is.... excellent teaching even for those who don't have in-depth knowledge LOL
Thanks a lot for this excellent tutorial ❤
Epic tutorial thanks mate!
Thanks great tutorial
thanks for your explanation!
Thanks
Thank you SO MUCH for the explanation of what is actually happening when u click or button or move a slider. I still dont really understand how stop at affects and image though. I get that it stops influencing at x percent of img gen but thats not very helpful as that doesnt tell me how its gna affect the img. What is the provess of how the img is gen’d? What gens first and what gens later so i can think concretely about what happens each percent step of the generation? What details are added in the first 10 percent vs the last 10%? For ex
Hey Joanna, thank you for the great question! The "Stop At" slider controls how far the AI goes in refining the image. Early steps (like the first 10%) focus on big shapes and overall composition. Midway steps add details like textures and shading, while the last 10% sharpens fine details and polishes the image. Stopping early means you'll get a less refined, more abstract result, while letting it go further creates a more detailed image. Hope this clears it up!
This article goes more in-depth into the entire process:
www.altexsoft.com/blog/ai-image-generation/
You are amazing
Beautiful Tutorial!!! Is it possible to make a tutorial on Fooocus on how to turn a picture to art. Example: A picture of a dog to watercolor style.
Have a look at the discussion of the Github link I shared.
Here is an example comment:
github.com/lllyasviel/Fooocus/discussions/557#discussioncomment-7226225
If you scroll further up & down you will see other people having similar questions and just see what they tried out.
In your case, if you want to have a dog in watercolor style, use 1 image of a dog and 1 image of a dog in watercolor style.
Just search on Google for an image you want your end result to look like, save it to your computer and just use that as input.
Then, play around with the stop at & weight of both images.
Start with both settings as you can see the person used in the link I mentioned above.
So, PyraCanny for your dog image, with stop-at at 0.4 & weight at 1.
And ImagePrompt for your dog in waterstyle color with the same stop-at & weight as the other image.
Regarding your styles in the panel on the right, just keep FooocusV2 checked.
Let me know if that helps?
Looking for a fresh take on image creation? Explore Image to Image in Focus and discover how AI can help you generate unique and visually appealing images. #AI #ImageEditing #Innovation #SmythOS
Merci, il manque énormement de tutoriel récent de sur Fooocus
Notre plaisir 😊
I am curious about the export resolution of the images, thank you so much!
There are actually quite a few different resolutions/aspect-ratios you can choose from.
For this, just tick the Advanced option below the text box for the prompt and go to the Settings tab.
There you can find for example:
1536x640
1664x576
1728x576
If you want to have an even bigger resolution, you can just drag and drop your desired input image into the upscale box of Fooocus and then just upscale the image.
If you don’t know how the upscaling works, we have a simple video about that on our UA-cam channel.
Let me know if that helps.
What is the best checkpoint for human realistic image, for good quality
Thanks
There is no "best" checkpoint per se. It all really depends on what you want to achieve.
The default settings of Fooocus already give some great results to achieve human realistic images.
Otherwise, you can try using JuggernautX as base model or instead of opening up Fooocuse normally with run.bat, open run_realistic.bat
Let us know if that helps :)
@OfficialGTDuo i try it but ondy portrait for face is good quality, when make full body image then always blurry, i try both realistic and run.bat, even i try upscaling, image size go to 25mb but image still blurry, skin texture like hand painting
Have you tried Fooocus inpaint in that case?
We have a tutorial about how to use Fooocus for improving details in an image like the face:
ua-cam.com/video/P4N1HRWj2Jo/v-deo.html
Also, you can have a look at this Github discussion where there are some more examples on how to improve faces (scroll down to the section "Advanced: Detail Fix (face, hand, eyes, ...)"):
github.com/lllyasviel/Fooocus/discussions/414
You will be astonished by the results you get when using the inpaint function.
Let us know if that helps.
When I hit "generate' nothing happens.
Clicking the Stop & Skip buttons does nothing.
I have to reload the page.
I'm using online version of FOOOCUS.
What do you mean with online version? You mean on Google Colab?
@@OfficialGTDuo
It's called 'rundiffusion'.
Cloud version of Stable Diffusion.
For users who don't have sufficient GPU power on their own devices.
@@OfficialGTDuo
It's a cloud version called 'rundiffusion'.
For those who don't have sufficient GPU power.
With such a cloud service it is possible that you have too little storage or GPU power on lower price tiers and might have to upgrade to a higher pricing tier to get it to work properly
@@OfficialGTDuo
8gb ram
Intel HD Graphics 4600
But yeah.....I def will upgrade.
How to open fooocus?
In the folder you downloaded for Fooocus there should be a file called run.bat which you just have to open.
It will open up a command prompt window, which does some setup and then it will automatically open up your default browser where you will see the UI of Fooocus.
Alternatively just check out our video which shows how to install and start Fooocus here:
ua-cam.com/video/s10LL9cUQmQ/v-deo.html