Create ANY Image with Fooocus AI
Вставка
- Опубліковано 15 вер 2023
- Learn how to easily generate whatever images you like using AI with the Fooocus version of Stable Diffusion
Requirements: Windows, NVidia Graphics Card with at least 4 Gigs of VRAM, about 18 Gigs of Hard Drive Space
Fooocus Github:
github.com/lllyasviel/Fooocus
ZIP Free, Open Source ZIP Utility:
www.7-zip.org/download.html
CivitAI Emma Watson LoRA:
civitai.com/models/123328/emm...
The best tutorial I have seen about Fooocus. Brilliant! 😀
Fantastic bud, keep it up!
Best tutorial I have come across😍
Very helpful
Thanks
Thank you so much!! Do I need to download nvidia graphics card drivers or are they already pre-installed in a PC?
At least some version would be installed otherwise your graphics wouldn’t work at all. However how recent it is another matter. I would use the Windows System Information tool and find your graphics card model and then download the latest drivers directly from Nvidia or download their GeForce Experience software which should auto detect your card and install the latest drivers
Thanks for the tutorial, huge help! How do I open fooocus in the browser again after I closed it? Do i just run the run.bat again?
You can the copy the link from the DOS window (scroll up if needed). I believe it is 127.0.0.1:7860 (from memory). That is the local host address and port 7860 (it basically starts it's own mini web server on your PC for local access, pretty common with Stable Diffusion interfaces)
Hey thank you for this amazing tutorial.
I have a question if I download it all on other pc and copy it to my other pc with all the models downloaded will I still able to run it?
Like the internet in my office is high speed so I want to avoid downloading time at my home.
Is it possible if i just download it on my office pc and copy it to my home pc? Will it still requires to download something?
When you pick certain Speed options or start with the different batch files or do something like in painting, it downloads the needed files at that time. However once those files are downloaded for what you want it to do, you should be able to just copy the folder and move it to another PC
@@holidayeffects3933 you mean I have to first run and download all the speed and other features that I will use in the 1st computer and then I can copy it to my other pc right?
Will it save my time for that 18 gigs?
@@arxkhan6280 Yes that is correct. Whatever feature is working is already downloaded and should work fine when copied
Hello mate. Can you please describe how can we change model in foooocus ? Suppose i like some model from tensor art and want similar there ( without faceswap ) how can i do it ?
@@thribsilva I tried running locally , unfortunately it requires high end devices 🥲
hey, sorry for bothering, I saw that u do video on fooocus, so I wanted to ask you one question. I generated a image of a girl and now I want to male AI instagram profile of her-make a lot of photos of the same model. I did it once and now when I want to make a photo of her - for example: make her pose for a picture on a crowded street or idc, something else. What should I use in the input image and how, thank you in advance.
release colab tutorial for civit models
Hi, thanks for letting me in on this. Are you able to tell me why I cant generate anything though? It takes about 12 mins then comes up with 'Error timeout' Thanks
Check the black, DOS window running behind, it should show the error message. You do need an NVIDIA GPU with at least 4 gigs of VRAM, so that may be the issue.
@@holidayeffects3933 HI, thanks for replying. I sorted it I had to re-install it for some reason, but it works now. I have a 3090 RTX so I'm good there, could be better, but could also be far worse lol. Thanks again. 🙂
Awesome! I guess something went wrong in the first installation. I am using a laptop 3080 with 8 Gigs VRAM so yours should be even faster (mine takes about 30-40 seconds each). I normally set it to gen 4 at a time while I am doing something else so I just check on the results after a minute or two.
@@holidayeffects3933 30-40 isn't to bad really. I find mine takes about 14 secs per image. To think this should hopefully continue to get better and better like all AI in the coming months and years is amazing. Once again thanks for sharing this with me.
just update your grapgic card driver (nvidia or what ever it is )
How can I do the Emma Watson part but with an image of my own?
I want to create a girl based on a photo of another girl
Hi thanks for the video. To Uninstall, do I just delete the folder?
Yes, to my knowledge everything is setup and stored in the Fooocus folder
Nope if you used terminal install there are tons of dependancies it installs.
How do you download the lycoris models?
When I press the Generate button I get Error. Nothing more. Have you any idea?
I got it but the pictures get corrupt or just turns black. Any idea on what i can do?
Can I use a picture and turn it into a cartoon?
How long does it take for you to generate on image? I'm on a macbook m1 pro and it takes more than 30 mins.
Right now Fooocus is really only for PCs with an NVidia graphics card running Linux or Windows. On a laptop 3080 it takes about 30 seconds per image. Unfortunately only NVidia has had long term support for AI accelerations and AMD ROCm and Apple Silicon are trying to get there but the software is mostly geared towards the NVidia platform right now. You might want to check out Automatic 1111 or Comfy UI for Mac support. They tend to be more bleeding edge on the technology while Fooocus is, er, focused on an easy user experience
@@holidayeffects3933 thanks for the response. I will check comfyui and automatic1111
Hi. I couldn't get Focus workin on my laptop no matter what. I get paused error on the console. I downloaded Pinokio and installed from there but this time it gave Cuda memory error. I have Nvidia gtx 1650 4 gigs ram, amd ryzen 5000 series cpu and 6 gb of ram. Is this not enough to get it work? I got ComfyUI work at 512*512. But the interface is terrible. Any suggestions about getting Focus work?
buy better videocard, 8 gb is a minimum
No. On their page it says that Fooocus can be runned by 4 gb vram.@@bazarbekovic
Hi I need some help installing this. I have downloaded and unzipped the file, but when running the run.bat file the command prompt comes up letting you know to hit any key to start the download, however any key I hit closes the command prompt and nothing happens.
I have found huggingface has been getting stuck on downloads lately. Make sure you are executing the "run.bat" in the root and if the download gets stuck just close the DOS box and try again.
do we need to tap the other run bat file or it will automatic? and how long is the safe time to wait?@@holidayeffects3933
Is there anyway I can add my own image for a baseline instead of someone elses photo?
Yes just click the "Input image" checkbox in the lower left corner and then go to the Image Prompt tab to upload up to 4 reference images
@@holidayeffects3933 I thought that is what the checkbox was for. Just wanted to make sure. Getting into coding after working in the data center as a data center tech since 08. Working my way through CCNA, but that's really just to get the cert. Figured I'd jump into a project that would force me to learn code. So I am working on fine tuning a StyleGAN. In the meantime though, I would love to also learn these platforms to keep me motivated. Thanks for the quick response!
Am I the the only one that have this error while executing the run.bat file ? ConnectionResetError: [WinError 10054] I'm totally unable to install it. And I don't find any information on this error on the web.
When you run Fooocus for the first time (and if it detects any updates) it downloads a bunch of models and settings. It is very likely and antivirus or firewall setting is preventing this from occurring, so I would check those.
Will this destroy my pc
Like I make a complex one or something
And
Will it take away ram from pc
(Please lmk as soon as possible)
Thank you
I have generated hundreds of images on a laptop with an RTX 3080. If I only generate one picture at a time the fan doesn't even spin up. It is certainly no more demanding than a lot of AAA games. Also there is really no such thing as a more complex stable diffusion image. Images take 25 to 30 seconds to generate no matter what on my hardware. You do need an NVidia GPU with at least 4 Gigs of VRAM and 8 Gigs of RAM in your computer as a minimum and the closer to the minimum the slower pictures will render BUT it is no different than running a challenging game on your hardware.
I have RTX 2060 and 16GB RAM. I render those in background without problem. I just found that you need to let everything load in the cmd window till the very end then it works waaaay faster. At start I've been turning it on and start render right after it starts in browser. I accidentally waited more one time nad noticed it works waay faster and can work in background now instead of making whole machine freeze and unable to do anything else than this :) Great tool btw
@@AI-Stories-Fun Good info - I also found the program regularly updates (with big downloads) and if you use a new feature (like Image to image) it downloads large files on it's first use. After that it runs smoothly. When in doubt check the DOS window and see if it is downloading/updating
@@holidayeffects3933 Yeah. I've noticed that too. It's not downloading inpaint tools until you use them. It's really great that it updates so often too. Good to see it's evolving fast. I've downloaded couple LORAs and MRE version. And having fun with it.
Without nvidia graphics card work?
Fooocus has added instructions for AMD GPUs and Macs. Scroll down the GitHub page to the section that applies to you like “Windows (AMD GPUs)” for instructions in how to set it up for non Nvidia GPUS
I need help my downloaded lora is not working.
When downloading from a site like CivitAI, make sure the Lora you are downloading is for base model "SDXL 1.0" and the type is "Lora" and not a checkpoint, vae, etc. Place the downloaded file in to the "models/lora" pre-made subdirectory under the Fooocus directory. Finally check for keywords in the description or sidebar. Many loras won't activate without the keyword. You can see all of that in the sidebar and even filter for just SDXL in Civitai. Then in Fooocus, click off everything under "Style" tab and then in the Model tab select the lora and set the strength typically to somewhere between .7 and 1 and make sure to include at least one keyword in the prompt
Mine says error en a red oval on the page any ideas ?
Disregard I needed to update my Nvidia driver
@@dylanharo7277 Glad you got it figured out :). One other tip is if you want to use a new feature like inpainting or change speed settings it needs to download things which can take some time and simply won't work if you are offline. Make sure you let it download everything it needs and then you can go offline with it.
bonjour cela ne marche pas il me marque C:\Users\user\AppData\Local\Temp\7zOC10EDB04>pause
Appuyez sur une touche pour continuer...
Make sure you have enough space on the C: drive - it downloads many large files
"fishes" = the fish
LOL Totally agree, unfortunately there is a loose wire somewhere between my brain and my mouth that shorts out occasionally
Why sometimes it cuts the head?
If that happens you need to describe the pose like "portrait", "full body", "face", etc. as part of the prompt such as "A full body photograph of an Asian woman wearing an ornate, traditional gown"
@@holidayeffects3933 I speak spanish. Maybe is it better to write the prompts in English?
I have only tried English prompts so I do not know how well other languages are supported. However I do know all of the styles internally are in English so I suspect English will give you much better results.
Thanks!@@holidayeffects3933
I'm testing Fooocus, but it's melting my computer (32GB RAM, Ryzen 9 5900, Radeon 6800XT). Temperatures are off the charts, I can boil water on my GPU. There were a few times when the computer just shut down and I wondered if it would restart.
Any ideas on how I can slow down Fooocus to render a little longer but not burn my computer to the ground? :)
Better check your airflow, cooler and chassis intakes.
It's pretty surely a cooling problem, not a workload problem.
I agree with @UnifiedFriends, your PC cooling should be able to handle the load so something is wrong. I typically use a laptop (rarely great at cooling) with Fooocus and although after a few pics the fan spin up to high but nothing that a AAA game or blender render wouldn’t do. Some cases are horrible at air flow you might even try it with the side(s) off and see if that helps. Also make sure your fan curves aren’t set too low and no cables might be restricting airflow. Something is defiantly up with your cooling though.
@@holidayeffects3933 Thanks for responses :). Cooling and chasis was ok, I rendered in Blender before and I hadn't got any problems. That's why I was surprised with issue with Fooocus and sdxl.
But, I undervoltage just a bit my GPU and it resulted in low temps for more than 30C X_x
@@JieTie I am glad you got it under control with an undervolt (good thinking). I am still mystified however because, to my knowledge, stable should just be a standard GPU load although I know it is just using specific functions of the GPU. I will do some research and see if this has happened to anyone else. Very curious at this point.
@@holidayeffects3933 Thanks for showing interest :). As I mentioned; I've baked many times in Blender and renders where sometimes even 20-30min. In Fooocuse, just after... maybe 30 seconds GPU was at 90C. I tried maxing fans but they couldn't keep up.
is copyrighted ?
It is open source & free under the GNU Public License
You can not turn text into an image. Because as the saying goes, you will need 1000 words to describe a picture. This tool is another tool for graphic artists.
lol what...
I would be careful of downloading any of these FREE AI tools because many of them are full of nasty virus . Especially the ones that are clothed in an EXE format . I have tried to get fooocus going on my computer over 10 different times running on 12 gig memory and 4 gig gpu on Windows 10 . The best result I achieved was that the interface came up and I entered the prompt and it went through the 30 steps to generate the image but when it hit 30 the image disappered and the program hung , Just tired of trying to make it work . Thanks for the share. :O)
Sorry it isn't working, I suspect it is running out of video RAM. Try the "Fast" setting which has fewer generation steps. On the safety issue, Fooocus is a well respected project that is frequently updated, open source and the main developer is active in the project. Although nothing on the Internet should be considered 100% safe, I have found no reason to suspect this project at least
@@holidayeffects3933 does it actually take like 4 hours to unzip or am i doing something wrong?
Bro something's wrong on your side. Stable Diffusion is free of viruses. Many respected people have given it a nod