Overlooked Fooocus Tips and Tricks (and AI Hands)
HTML-код
- Опубликовано: 7 июн 2024
- Master Fooocus and Stable Diffusion for Creative Image Generation! 🎨 In this Stable Diffusion video I go over some tips and tricks and even dive into AI hands a little bit. Unlock your creativity with Fooocus and Stable Diffusion! #Fooocus #StableDiffusion #DigitalArt #ImageManipulation #CreativeTools
This is part of my ongoing series of videos on installing and using Fooocus for Stable Diffusion. Fooocus is a great interface for using Stable Diffusion and is very easy to get started. It also will run on systems with only 4GB VRAM.
If you find the video useful and would like to tip, RUclips does have the Super Thanks option. Those are greatly appreciated so I can have my coffee!!
☕ BUY ME A COFFEE: ko-fi.com/kleebztech
Also I am a verified creator with Brave and do accept BAT for users of the Brave browser.
Download Fooocus: github.com/lllyasviel/Fooocus
Driftwood Art LoRA: civitai.com/models/281998/dri...
How to install Fooocus video: • Stable Diffusion - Dow...
Fooocus Beginners Guide: • Super Easy Stable Diff...
Fooocus Upscales and Variations: • Stable Diffusion - Foo...
Checkpoints and LoRAs the Basics: • Stable Diffusion - Che...
Fooocus Poses and FaceSwap - Image Prompts: • Stable Diffusion - Pos...
Zoom with Outpainting: • Stable Diffusion - Zoo...
Inpainting with Fooocus - Don't Regenerate, Fix!: • Stable Diffusion - Inp...
Background Replacement with Fooocus: • Stable Diffusion - Bac...
00:00 Intro
00:45 Prompt Weights and Weight Keyboard Shortcut
05:45 Dark Mode
06:50 Style Search and Style Tips
12:43 Log and getting settings from previous images
16:53 AI Hands and ways to improve
22:10 Creating graphical text 
Наука
![Dimension 20: Never Stop Blowing Up Trailer [Dropout Exclusive Series]](http://i.ytimg.com/vi/xYeXpbHnGEM/mqdefault.jpg)
If you find the video useful and would like to tip you can buy me a cup of electricity for all these image generations. It is greatly appreciated! ko-fi.com/kleebztech
Highlight crtl-up/down arrows, OMG Yes!
That was my response. And it has been there for a few months without me knowing.
3:10 OMG thank you soooo much!! I had NO CLUE could do that, or that Fooocus could do weight for that matter!
Turns out when you want to reselect it you only have to click somewhere inside the parentheses on the word and not select the whole thing.
This is by far the most informative Fooocus tutorial I've ever seen.
True
Please keep making more of these videos. Thanks
More to come!
Awesome tutorials
Seriously, dude. Your tips are awesome. And thank you for answering me the other day about the adding a lora style to an existing image. I am not getting it right yet but I hope I will eventually. Keep it up!
Thanks for this, bro. I will definitely be putting these to use in my next generations.
Hello! I've finally caught up with the tutorial videos! Like the others, this one is also perfect and very well explained. The tip about adjusting the prompt weight using ctrl+up/down is a godsend. I can't wait for the next videos :)
Thank you!!
I have learned many useful tips and how-to's from watching your videos. Thank you for sharing your knowledge.
Glad to help
Great work dude! A small tip for the dark mode. You can get the dark UI mode by adding this in your .bat file: --preset realistic --theme dark
Good call! Missed seeing that flag. But then again I was not looking. lol. Of course I think 😀everyone should just be using dark mode for everything.
A dark mode discussion? I see you are men of culture. 🙏🏽
Good stuff. Thanks.
Loving these videos. ❤
Glad you like them!
Hi, thanks for doing these videos.
Just a little tip for you also if you'd like. Once you've done the highlight and weight sequence. If you want to go back and change it, you don't have to highlight again, just click inside the brackets and use the Ctrl up/down again. Hope that helps.
Excellent. 👍
great video, thanks
Glad you liked it!
Thanks, keep up the good work!
Thanks, will do!
another great video!
I can't thank you enough!
Thanks again!
As others have mentioned in the comments, when you go back to adjust the weight you do not need to highlight the text. You only need to click and place the cursor inside the text and you can adjust it. And there is a dark mode option if you use the CMD flag --theme dark if you want to go that route with the .bat file. But I prefer dark everything using the browser.
well, you kinda do need to highlight the text if it's multiple words.
@@mick7727 actually once you have added the weight you shouldn't need to highlight multiple words. Should be able to just click anywhere within those words
Version Fooocus 2.3.1 in Easy setup, you can select dark and other options
Thanks Sir
Welcome
I downloaded Fooocus a few days back, and your tutorial is very informative and interesting; saving me a lot of time on the learning curve.
Glad it was helpful!
Thx for the tutorial kleebz! Btw can you make a tutorial on how to get skin moles, pimples and other facial defects in portrait photos to get more realistic touch
You are welcome! As for the video on that stuff, I will have to keep it in mind because I would have to experiment with that since I have not really looked into it. I wonder if there might be some loras that would help.
@@KLEEBZTECH yes you could even try with latest models or may be by increasing weight..
Adding ("Happy Birthday" text) to the prompt may help. Put the words you want to see in text within Quotations and add text after
Usually end up with it adding happy birthday somewhere else. It doesn't see that text as text. So it adds more.
👍👍👍👍
nice video. u can explain how make thist video at 7:50 ...
u can make a tutorials ? TY
Do you have a tutorial for how you created your lora for your goat? I want to make a lora for a consistent character
I do not. I used an online service at the time that automated it. That no longer exists. I have been experimenting with doing them myself but the amount of time it takes and my limited experience at the moment I just don't feel I would be giving good instructions yet. But if you do come across a good tutorial I would love to know. I have been experimenting with OneTrainer and did get decent results on one for a friend so far. But others have been not so good results. I do know that your source images and labeling will be a huge factor. So I would start trying to build up a decent amount of good images with variety.
hey man, how can i make the background clear and not blurry ?
Are you talking about the background being blurred with things like portrait shots? If so you would need to work with your prompt to try to limit that. Remove things like portrait if using that. Also could try describing the background in detail. I have not really tried doing it so I can't say exactly what might work and every model will react differently.
Do you know if it's possible to create textures? Textures like for 3D. By looking at your videos and playing around with the software it seems like it's not something that you can just click but if you put texture map in prompt it helps a little bit. But is there a way to make it create uniformly lit symmetrical textures that are almost perfect for use?
Not that I am aware of. I might play around with that idea and see what is possible.
@@KLEEBZTECH ok. Thank you for the reply. Also, that advice to turn off all of the styles that you always give- very valuable one.
What graphic card U use? RTX 2060 should be enough for fast image creation?
I am using a RTX 3070. I am not sure what you will get for performance from a 2060 but imagine it should not be terrible if that is what you have.
thanks! just a little question: Why do I get 4 different images, using the same seed and the same prompt and not changing any parameters at all?
It uses the same seed for the first image. Then it sequentially goes to the next one for each subsequent image
@@KLEEBZTECH thanks a lot!
Great video, very usefull. You know, i still trying to inpaint clothes from other photos, but still unsuccessfully. Any advice? Thanks!
This is something I'm going to be playing around with this week.
Are you looking to get it exactly the same or just very similar?
@@KLEEBZTECH it should be exactly the same, because the idea Is to show the clothes in diferent people. When i tested using inpaint selecting only the clothes was that fooocus modified the clothes and the result was something different than I expected.
The only way that I could think of at this time is to inpaint all but the clothing and have Fooocus fill in a new person using the original image of the clothing. Of course that would be very limiting to just whatever the pose was and if Fooocus would do a good job.
@@KLEEBZTECH great, I'll try out. So can you tell me the set up you would use for this case? inpaint or outpaint + image prompt swap face?
For nvidia gpu whats the better driver for SD game ready or the Studio driver?
In the past I tried both and saw no difference. I use the game ready since I also do gaming.
@@KLEEBZTECH thx
Hello, i love your guides, but im using web google colab version and i have a problem, mby u can help me..
In my situastion im trying to use "mix image prompt and inpaint " in options advanced, control.
i alredy have 2 prepared photos. One with face i wanna use to faceSWAP and second where i have character to use in on with just inpaint face mask.
Im putting character in inpaint, masking face , then other picture in faceswap. and generate , and 100% time im gettin error, every try.
Mby u can suggest smfn i can do. Mby somehow load inpaint picture before ?
What are you getting for an error? I have a feeling that it may be partly that on the free colab it can have difficulty with some things. Might be hitting the limit on resources as mentioned here: github.com/lllyasviel/Fooocus/discussions/1318
@@KLEEBZTECH resources limit i think.
is great video thx but Can you do something like this : It's a close-up shot of an eye, reflecting the scene in its iris.
The eye has detailed textures, with visible eyelashes and skin texture around it.
In the reflection in the eye, there's a person riding a green Kawasaki Ninja on the road.
I can get it mostly but hard to see the actual reflection. Midjourney got it easy. I might play around with the idea some more.
All those aspect ratios and not one with 16:9, which is the most common ratio...?
You could create your own with a custom preset of in the config file or there is the option to override those in the debug section as well (Forced Overwrite of Generating Width and Forced Overwrite of Generating Height). I usually just go with one of the closer ones.
They went with ones that gave the best result from my understanding.
1:50 the seed is the last 3 images generated? and why you can change the seed number?
Once I generated the first set of images I unchecked the random box so I could use the same seed so I could show the difference. When you uncheck the random box it leaves the seed for the first image in the box. All the seeds in a batch are sequential so it will use the same 4 seeds since I generated 4 previously.
@@KLEEBZTECH and that means it bases on the same images ? Im downloading Fooocus MRE to try image-2-image function among other things... Does somebody knows how it works?
Using the same seed with all the same settings and prompt should get the same image. So using the same seed when making slight changes can help you see what impact those changes have.
As for the MRE fork I am not sure how up to date that one is. The standard Fooocus does have the image prompts and such.
class, handle and doors knobs on kitchen moving and in wrong position
I have not found a good solution yet. Have tried a couple Loras with little success.
What's with the same framing on ALL of these pictures?!
Not sure what you are asking?
It's all close up macro shots. I have Fooocus installed with like 40Gb of LORAS and I can't get anything full body!@@KLEEBZTECH
I do have a video that should help with that here. Much has to do with aspect ratios and how you prompt: ruclips.net/video/YODdFMlZz2Q/видео.html
Actually if you edit the run.bat file like so: .\python_embeded\python.exe -s Fooocus\entry_with_update.py --theme dark
pause .....you can have the dark theme fire up by default
Yes. I had missed that flag. Although I prefer to do with the browser myself. Dark mode for everything.
@@KLEEBZTECH saves a lot of energy I Believe ( besides ones eyes 👀) thanks for the great video
Hands are really a painful point. I think maybe good starting with an Image Prompt of some kind and than refine and refine the hands with different techniques. As always more good is the input than better will be the output
Best thing is to avoid hands. lol. Or also hands that are not doing anything complex like holding something.
you can just write in an argument --darkmode
The left hand fingers of the woman in the blue dress don't look natural either. They're a kind of missing phalangeal joints.
I don't understand how developers of the program can't simply train AI specifically on the hands. Maybe training should be gravitated around the anatomy of drawing, so the bone structure can be understood better. After all drawing anatomy tackles exactly the structural build of the human body.
I think we can expect that sort of thing in the near future once they have a better understanding of how to train it better. The big issue is the fact that it is trained on low resolution images. Take any image you have in your personal photos and reduce the resolution down and see what the hands look like. Once they can train them to understand other Concepts like anatomy and everything together with the image generation would be helpful at some point. And all that I think will come soon.
@@KLEEBZTECH Inevitably it will come. Meanwhile, I will have to use ComfyUI. Cheers!