Fooocus - Add Custom Settings/Resolutions/& Even a Turbo Preset!
HTML-код
- Опубликовано: 19 июн 2024
- Setup Fooocus how you want with an SDXL Turbo Preset, Custom Resolutions, Preferred Default Styles, File Paths, and more!
Fooocus is Stable Diffusion on easy mode and this guide will help make it even easier and more efficient. Set it up how you want and create easy presets of any settings you wish.
If you have yet to Install Fooocus, Stable Diffusion's easiest AI Image Generator. Check out this video here: • Fooocus Tutorial - Sta...
DreamShaper XL: civitai.com/models/112902?mod...
------
00:00 Intro
00:22 Config Customization
04:19 Getting An Error?
04:50 SDXL Preset
------ 
Наука
Fooocus Download Page: github.com/lllyasviel/Fooocus
Will this work for Colab as well?
@@osnapproductions1962 I haven't used the colab version too much, but I believe changing the settings should work just the same. However the free colab version still has limited vram usage. So anything that goes over that will still end with an error.
How do I use multiple model folder?
Thank you. You are the first one who did the Fooocus Turbo Preset tutorial on RUclips. I have been looking for a long time. I read on some forums, but I did not understand. You have made it easy to understand and apply.
No problem! I do try to find useful things on the discussion page that I don't see people talking about. An SDXL Turbo Preset was a big request.
Thank you for your tutorials. I have gained a lot. I look forward to more exciting tutorials from you.
Awesome tutorial as always, thank you!
Welcome!
Thanks, great Fooocus Inside
Thanks for the help, started with me not knowing how to change resolution presets and ended with a basic understanding how to set up new models on my own
thank you very much!
you saved me a lot of time!
Thanks alot man!!!
Muchas gracias compañero! :D
Thank you!
Welcome!
thank you ! thank you ! thank you !!!
You are welcome!
you briefly covered adding a line for embedding and it looked similar to how you might add a lora via prompt but I also noticed there is an embedding category on civitai so there must be some kind of difference between a lora and an embedding. I would like to know about advanced prompting, weighting individual words all of the open and close tags () this:that:1 etc, positioning prompt words having an effect on image generation. basically just an advanced prompt guide would be cool. thanks for the vids
Great info! I didn't realize you could render higher custom rez images than the default. I'll have to give that a try. I'm guessing there are more artifacts and deformities the higher you go, and also the need for a high end pc/gpu? Thanks for this guide, all be saving this vid!
Yes on all counts, more VRam, more artifacts. But the option is there for those who really want to use a different resolution.
Great video! I hope IIIyasviel adds easier ways to adding presets similar to the "speed, quality, extra speed", so it's easier to change. A pose and depth control net would be the icing on the cake. Having easier access to combining image prompt and inpaint or upscale/variation would be great as well.
There's the fork "Fooocus Control", but I've heard it changes things unnecessarily, instead of being just extra controlnets on top of the default fooocus features.
Yeah I believe a big reason for not including an SDXL turbo preset was the "non-commercial" license issue.
I wanted to like Fooocus Control but my results were all over the place. And it appears to be abandoned or at least not updated in many months.
A lot of the issues are just the lack of developed controlnets for SDXL itself. But I would love to see an openpose/dwpose in Fooocus someday.
Hi, is there a way to get a middle ground betweend "Vary (Subtle)" and "Vary (Strong)"? Or is it in some way restricted?
Yes, you have to dig into the debug menu but you can control the denoise. In debug mode>Debug tools tab there is a setting near the bottom called (Forced Overwrite of Denoising Strength of "Vary") you can change this to 0 - 1.0
I believe the "subtle" denoise it 0.5 and the "strong" is 0.85 to get an idea.
Also it wont matter what "vary" you use once change the denoise setting. they both will be the same.
Hello, thank you for your video :), I have a question :p. I would like the images to be saved in the ouputs folder by default EXCEPT when I click on the circle on the right on Turbo I would like in this case the images to be saved in another folder and when I click on another preset we return in ouputs, I hope I explained well. Thanking you.
I understand, you want to click on a preset and have your outputs go do a different folder with each preset.
you CAN change the output folder of a preset. But It doesn't seem like the output folder will change when switching between presets inside of fooocus (at least from my testing). If you start fooocus from that preset all images will go to that specified folder but wont change between presets when switching inside of fooocus.
@@JumpIntoAI thank you for your answer, that's what I thought :( , too bad, if necessary I would restart fooocus :) otherwise I succeeded by creating a 2nd temporary folder in my file \Local\\Temp\\fooocus2", I also change the output. Then I start the default file and then the one I modified, so I have 2 windows and it works, of course I only use one at a time and I switch from 'one to the other :) (sorry for my English, I speak French and I use google translate :))
125s/it
can someone help please?
Is there an option for using jpg as the default output format instead of png ?
yes, it was recently updated that you can now save in png, jpeg and webp.
That option is in the settings tab between the image number and the negative prompt.
Is v2.1 Turbo DPM++ SDE a newer and well-working version of DreamShaper as well? And better than sd_xl_turbo_1.0.safetensors?
Yes the v2 Dreamshaper is the newest version "released just after i posted this" should work just the same.
@@JumpIntoAI Thank you.
This config_modification_tutorial file does not exist anymore? Weird
but why when i added another aspect ratio like 9:16 or another ratios added it gives me the result of the photo bad ?
These models are trained on specific aspect ratios. Using custom ones will give worse results more often. It is recommended to use the defaults but you can sometimes get away with custom resolutions.
I coudn't find config file in automatic1111 which one has path ...... can you help ??!!!
The file is in the Fooocus folder, config.txt. If you have "automatic 1111" installed already and want to share the models/checkpoints from that with "Fooocus" so you don't have duplicates on your drive. You just want to put the model paths from automatic1111 into the fooocus config.txt.
for example my main checkpoints are installed at "X:\a1111\stable-diffusion-webui\models\Stable-diffusion" so that's what I want to put into the fooocus config.txt line where it says "path_checkpoints" but for it to work i have to add double backslashes to the lines. so it would actually look like this in the file "X:\\a1111\\stable-diffusion-webui\\models\\Stable-diffusion"
If you are trying to point your automatic 1111 folders somewhere else I don't think that works the same way. Last I checked you need to edit the webui-user.bat and add a command line argument. such as
set COMMANDLINE_ARGS= --ckpt-dir C:\checkpoints
That would work If you had all your model/checkpoints in c:\checkpoints
Sorry if that was more confusing than helpful!
Do you know why my Fooocus Interface is white and yours is dark? Is there a configuration to change it?
you can add "--theme dark" into your run.bat like this
.\python_embeded\python.exe -s Fooocus\entry_with_update.py --theme dark
Thank You@@JumpIntoAI
what should be performance option for custom turbo preset??
Leave it on quality or speed (you will get the same results with either) as a properly setup turbo preset will overwrite the settings that either of those use.
However, switching to extreme speed or lightning performance settings will override the Turbo preset and will probably give you ugly results if used with a Turbo Model.
how do I put it in cfg default "Advanced" checked & him options?
or developer debug mode checked
Error
File "C:\Fooocus_win64_2-0-50\Fooocus\modules\patch.py", line 475, in loader
raise ValueError(exp)
ValueError: [Errno 2] No such file or directory: 'D:\\sd\\models\\Stable-diffusion\\dreamshaperXL_turboDpmppSDE'
!!! D:\sd\models\Stable-diffusion\dreamshaperXL_turboDpmppSDE --I did everything as in your instructions, I don't understand why the error?
file not found, when you input the file name did you include .safetensor? "dreamshaperXL_turboDpmppSDE.safetensors"
Yes, I've tried it with others, it doesn't work, error!@@JumpIntoAI
is there an error to enable anyway - ValueError: No such file or directory: "D:\\sd\\models\\Stable-diffusion\\turbovisionxlSuperFastXLBasedOnNew_tvxlV431Bakedvae.safetensors"@@JumpIntoAI
Can you please make a video on how to get mid journey level details and quality in Fooocus, i mean Fooocus is not good at other things than generating pictures of people and womens in general
It does take a little more work but (depending what you are going for) a combination of LoRAs and prompts can get great results.
Why the f is it such a hassle to do something so seemingly simple?