Fooocus is great for "out of the box" experience, I do hiiiiiiiiighly suggest using the MRE version tho. If you struggle with XL, I can almost promise Fooocus will not be a problem. The longest wait is the first load.
@@Fanaz10 It's a fork of the original Fooocus which tends to have more features but do note that the original is usually the most up-to-date than forked version... Btw there are other forks as well.
Olivio, I have to thank you for broadening my horizons in terms of web UI. I discovered your channel a few days ago and it's helped me catch up on what I've missed over the past year. Your explanations and guidance are top-notch. Despite using a different setup/hardware, I was able to apply most of your tips to my Mac setup. Much appreciated!
10 times SLOWER on MINE - from normally 4 images with the same settings it takes 50 seconds - This Fooocus takes 5 minutes only for 2 images.... Disappointing
Great tool. And MRE version makes it even better if someone wants more control over it. I have real fun with this. Also it works superfast which is big plus.
hi, I wanted to ask you, if we already downloaded the basic foocus version, do we need to download entirely the foocus mre version in order to use it ? Or is there only a few files to download? I have a bad Internet connection so I wanted to know if there is a way to fasten the process, thank you in advance!
@@kevinytb6446 I guess that base yes. But models which are biggest part of it you don't. But MRE hasn't been updated for some time. I've came back to basic Fooocus couple days ago and I see it has changed a lot. New models used as base (Juggernaut v6), new toos. I use it again now. Especially after MRE started to work slower for some reason for me....
Lol. He doesn't mean the rendering is faster. He's comparing the work flow... How fast it is to get your work done, free from distractions due to the ultra clean and simple UI compared to the crazy messy complexity of the A1111's webui.
🎯 Key Takeaways for quick navigation: 01:37 🧩 The UI is quick, simplified, and concentrated on having fun with AI 01:37 🧩 The UI uses fewer resources and focuses on the art creation process 02:29 🔧 Users can specify the location of their already downloaded models in the user path config file 04:44 🎨 Users can choose from a wide range of pre-created styles or create their own for generating art 06:05 🔬 Users can select different models, such as chargernaut XL, from their model list for generating art
Installed this last week with the hope of using SDXL after watching a video where they mentioned 4GB vram was minimum. But it took me ages to generate a simple prompt so eventually had to delete it. SDXL is powerful so maybe it needs more vram.
@@CloudVideoYT Yes. Something is causing this issue. Also Fooocus has no interface controls for the image generations like SD. So its not worth it for me right now.
@@dreamzdziner8484 yeah I thought it same but actually fooocus is creating better image. You don't need to select multiple options to get better results. Fooocus have simple button option to choose from, that's why I like to try fooocus but unfortunately it won't run on 3060.
What no mention of Fooocus MRE??????? That is an IMHO much superior fork of Fooocus which brings into play some of the key features that Fooocus leaves out whilst retaining the default view simplicity and low VRAM requirements. For most people with a keen interest in generative art and a 4-20 G VRAM GPU it's likely to be the one to go for. I suspect that Olivio will do a review once he twigs it, it seems only to have been covered by some of the less popular YT bloggers following SD so far. The key problem is that SDXL only works slowly on GPUs with under say 24GG of VRAM. Fooocus works much faster on systems down to 4GB of VRAM, but it is very paired back with many of the A1111 features that intermediate users will miss. Comfy of course works much better with SDXL, but the whole node mode of operating seems too much of a learning curve for the average user.
That looks like a great tool for bashing out prompts to see what looks good, and I know whom I can recommend this to. The Seed thing is actually an issue for me as most of my work focuses on generating consistent characters without Lora. So in my Comfy workflow I have a face detailer node with a fixed Seed, and the seed on the Ksampler set to Increment.
I wanted to try this for my channel but SDXL really takes up too much resources that I couldn't even get it run sadly. The models look amazing.. Hopefully 1 day I will be able to try this
Man automatic 1111 got worse in mu opinion, sometimes i get cuda errors of memory out of nowere when 50 seconds after it created an image without problems.
How do you get the dark UI theme? Also is there a way to run it on my local home network so that I can use my laptop as a terminal to run it remotely on my desktop computer?
Год назад
Hi, great video about Fooocus. I already download it but not used since a long time and is now better, so thx for the video. One Question, you mapped only the models folder, why ? Is it problematic to map the others one at A1111, like the lora and uspscale folder too ? Thx
Thanks for another informative Vid! However I did make the suggested changes to the "user_path_config.txt" linking to the already downloaded SD checkpoints, but it seems to rewrite this file each time Fooocus is loaded with the run.bat file. Anyone else have this problem or fix?
@@speaktruthtopower3222 The first startup of WebUI even on my old i7 7700 released in 2017 and HDD (not SSD) takes 12 seconds, right now I measured this: from launching the startup .bat file to automatically opening a new tab with WebUI in the browser. Subsequent image generations will also be comparable in time. Obviously, a simpler and less functional tool without a couple of dozen plugins will be faster overall. It's silly to argue that. But if you use all the features of the A1111, you'll use it for both basic and complex tasks. There is no need to install other software in this case.
hmmm... Ich glaube fast die Pfade sehen so seltsam aus mit den doppelten \ weil die Verarbeitung intern sonst nicht funktioniert. \ wird in der Programmierung oft genutzt um Sonderzeichen die sonst andere Funktionen haben nutzen zu können. Nicht das es irgendwas daran ändern würde das sie vorhanden sind.
It's slow compare to automatic1111. I have 3060 12 gb, it takes 4+ mins to generate 1 image. I have installed couple time, changed os but still same issue.
It's impossible to use, I just installed it and it consumes my 16GB of RAM! I have never seen in my life a program devour all my RAM. The GPU was what should matter, but apparently it just uses all the RAM without doing anything, just opening the generated link. I can't have 40GB of RAM or more, 16GB is enough for everything here, it's never been a problem. I have a good video card and I can't use it because it sucks up all the RAM, I don't know what else to do.
I am running win 10 with gtx 745 gpu and 12 gigs of ram and I have tried and tried to get this to work with no luck . The interface shows up and it renders one picture and when it hits 30 the picture closes and the program hangs, Iam tired of it. :O)
Hello. Thanks for tge instructions. I downloaded, but by running it doesn't generate anything, error , error. And, i not found user_path_config file in folder. And one more, my laptop is on Radeon graphics, is it important? Pls help
nvidia still has driver issues and generating is very slow - had to revert to april drivers and work ok again. I wonder if this is also the issue with Foocus
I don't see the user_path_config file, do i need to create the file myself or maybe I need to run the bat file for the first time? edit: ah ok yeah i just had to run it for the first time, then it created the file for me.
did it have to download the xd model before it gives you the user_path_config.txt..im downloading the xd checkpoint it now and still no user_path_config.txt
@@hairy7653not sure. I already had the SDXL base and refiner models since I use them in A1111 so I just copied them over temporarily to Fooocus folder so it wouldn't have to download them, then i ran the bat, verified Fooocus was working, then i was able to edit the user config.
It's a nice UI overall, but the fact that it loads the models EACH TIME you make a generation make it really really slow, i can make dozen of generations on A1111 between 2 generations on Fooocus...
HI sir , I Have Almost 12 GB Of Ram Installed With Nvidia GTX 1650 But My Laptop Got Shutdown unexpectadley While Generating The Model. And Generating The Model Takes Too Long. Any Help would be Really Appreciable ?
Mine doesn't have the user option to add the location of the models as I'm using comfy ui too, And when I run it has network errors, so is there a fix for it ?
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? :)
Fooocus is great for "out of the box" experience, I do hiiiiiiiiighly suggest using the MRE version tho.
If you struggle with XL, I can almost promise Fooocus will not be a problem. The longest wait is the first load.
Yes, I like it also. And also the seed is random or fixed. And you can activate more the one style. And fooocus work with sdxl.
Very cool great for anyone that wants to use SD without the tinkering.
actually fooocus have most things already including conrolnet
I was waiting for you to make this video. Fooocus-MRE is amazing. Oh, wait, this is for base Fooocuss. You need to make one for Fooocus-MRE
what is MRE version?
@@Fanaz10 It's a fork of the original Fooocus which tends to have more features but do note that the original is usually the most up-to-date than forked version... Btw there are other forks as well.
Fooocus MRE is the Best with control net, loras, XL models, styles, etc etc etc
im surprised he did a video on the base version and not Fooocus-MRE.
Invoke is better
what is it?
Olivio, I have to thank you for broadening my horizons in terms of web UI. I discovered your channel a few days ago and it's helped me catch up on what I've missed over the past year.
Your explanations and guidance are top-notch. Despite using a different setup/hardware, I was able to apply most of your tips to my Mac setup.
Much appreciated!
To change the seed: you have to uncheck the random button whilst in advanced settings!
You saved me alot of time and money, thanks sir! =)
Never will be an A1111 killer in my opinion, but could be a great entry point for beginners since both share that gradio interface.
Yeah, not even close to an A1111 killer.
Make video on Fooocus-MRE !!!
what is it?
works on my potato PC - acer nitro 5 1050 gtx 4GB, really slow but i have no rush !!!, thank you so much for the video ! !!!
Thank you my friend to explain to me how to use it because I had this and now I'm using it it's really good I hope they improve it a little bit
Thank you for your video! It is 10times faster than A1111 on my machine 😊
Makes the generation process much more enjoyable 👍👍
Hi, sorry what is your GPU please? thank you
10 times SLOWER on MINE - from normally 4 images with the same settings it takes 50 seconds - This Fooocus takes 5 minutes only for 2 images.... Disappointing
Amazing new tool, thanks for sharing Olivio
Looks promising. A1111 is a bit overloaded.
Seed and CFG settings available in Fooocus-MRE version
you are the coolest dude I have a subscription for this video
thank you very much, I noticed it's already on Pinokio (which I also learned from you) so I gonna try install from there
Great tool. And MRE version makes it even better if someone wants more control over it. I have real fun with this. Also it works superfast which is big plus.
hi, I wanted to ask you, if we already downloaded the basic foocus version, do we need to download entirely the foocus mre version in order to use it ? Or is there only a few files to download? I have a bad Internet connection so I wanted to know if there is a way to fasten the process, thank you in advance!
@@kevinytb6446 I guess that base yes. But models which are biggest part of it you don't. But MRE hasn't been updated for some time. I've came back to basic Fooocus couple days ago and I see it has changed a lot. New models used as base (Juggernaut v6), new toos. I use it again now. Especially after MRE started to work slower for some reason for me....
@@AI-Stories-Fun ok ok 👌thank you for the answer
Foocus-mre has even better implementation ... Even something connected to noise and make even better pictures with less errors .
It says Fast on the title but you havent showed how faster it really is compared to A1111
Click bait
Lol. He doesn't mean the rendering is faster. He's comparing the work flow... How fast it is to get your work done, free from distractions due to the ultra clean and simple UI compared to the crazy messy complexity of the A1111's webui.
Yes, please show the comparision Olivio
@@ph0rd It also refers to the fact that it uses fewer resources, this helps those who only have 6GB Vram
It is faster
I am using Confy UI 4GB NVidea 3050, with no issues at all! One generation about 45 seconds, no Rifiner Models
🎯 Key Takeaways for quick navigation:
01:37 🧩 The UI is quick, simplified, and concentrated on having fun with AI
01:37 🧩 The UI uses fewer resources and focuses on the art creation process
02:29 🔧 Users can specify the location of their already downloaded models in the user path config file
04:44 🎨 Users can choose from a wide range of pre-created styles or create their own for generating art
06:05 🔬 Users can select different models, such as chargernaut XL, from their model list for generating art
What AI did you use for video chapter extraction?
@@Deffcolony HARPA AI
Super video. I like the embedded python. Almost makes it portable
YAY !
Nice, could you please explain why you write your prompt "(samurai_clothing)" instead of "(samurai clothing)"? Cheers
Thank you for showing this!!!!!!!!!!!!!!!!!!!!!!!!!! I Love You!!!!!!!!!!!!!
Samurai are not Chinese 😅
shhhhhhhh
Installed this last week with the hope of using SDXL after watching a video where they mentioned 4GB vram was minimum. But it took me ages to generate a simple prompt so eventually had to delete it. SDXL is powerful so maybe it needs more vram.
I have 12 gb & it takes more than 4 to 5 mins to generate 1 image. It looks like something is missed up. I don't know what's going on
@@CloudVideoYT Yes. Something is causing this issue. Also Fooocus
has no interface controls for the image generations like SD. So its not worth it for me right now.
@@dreamzdziner8484 yeah I thought it same but actually fooocus is creating better image. You don't need to select multiple options to get better results. Fooocus have simple button option to choose from, that's why I like to try fooocus but unfortunately it won't run on 3060.
Try without refiner when lacking VRAM or RAM.
@@AnandaFury I already try same issue
Fantastic video. Does it only do XL models or can you add in 1.5 base models as well?
Sdxl only
Fooocus is amazing! Slow on an AMD but blazing on my RTX...
What no mention of Fooocus MRE??????? That is an IMHO much superior fork of Fooocus which brings into play some of the key features that Fooocus leaves out whilst retaining the default view simplicity and low VRAM requirements. For most people with a keen interest in generative art and a 4-20 G VRAM GPU it's likely to be the one to go for. I suspect that Olivio will do a review once he twigs it, it seems only to have been covered by some of the less popular YT bloggers following SD so far. The key problem is that SDXL only works slowly on GPUs with under say 24GG of VRAM. Fooocus works much faster on systems down to 4GB of VRAM, but it is very paired back with many of the A1111 features that intermediate users will miss. Comfy of course works much better with SDXL, but the whole node mode of operating seems too much of a learning curve for the average user.
Looks VERY promising! A1111 doesn't need a killer, by the way - it already kills itself due to bad and unoptimized memory usage 😐
...and bad UI
Hmmm looks a lot like A1111 with a bunch of extensions installed... not a big fan of A1111 wrappers. They tent to fall behind A1111 quick.
Fooocus is based on ComfyUI with improvements like advanced AI prompting ( based on gpt2 ) under the hood.
It’s much better than that and has a lot of custom elements. Try it
I love how on the website they state that a MacOS version supporting AMD gpus is coming soon ;)
This looks good, more UI need direct API input ala Photoshop etc.
That looks like a great tool for bashing out prompts to see what looks good, and I know whom I can recommend this to. The Seed thing is actually an issue for me as most of my work focuses on generating consistent characters without Lora. So in my Comfy workflow I have a face detailer node with a fixed Seed, and the seed on the Ksampler set to Increment.
Are you saying we can’t have character consistency with fooocus?
Have a look of Fooocus MRE. Better fork 😊
Fooocus is for graphical AI what jamming is for music, just go with the flow.
I really hope this gets more development, I'm not a fan Comfyui and we need simpler alternatives.
I wanted to try this for my channel but SDXL really takes up too much resources that I couldn't even get it run sadly. The models look amazing.. Hopefully 1 day I will be able to try this
I get lots of errors with my 4090 when try image prompt or when I try to upscale. As if it runs out of memory, but it doesn't. Any ideas?
this is end of the picture world
Please help. My load time is 50 minutes and downloading focus took 4 hours. My pc and internet aren't bad either so I don't understand
I am definitely skeptical, but I guess a simple UI is good for people who don't want the ability for complete control.
good video thank you
Holy Dumpling this one is AMAZING!! Now i feel like i'm cheating at AI art now 😂😂😂
Fooocus is great.
Olivio, why Juggernaut XL V2? this is old
Easydiffusion has automatic and automatic internet server configuration and multi GPU support I’m curious 👀 about this
Thanks!
Man automatic 1111 got worse in mu opinion, sometimes i get cuda errors of memory out of nowere when 50 seconds after it created an image without problems.
How do you get the dark UI theme? Also is there a way to run it on my local home network so that I can use my laptop as a terminal to run it remotely on my desktop computer?
Hi, great video about Fooocus. I already download it but not used since a long time and is now better, so thx for the video. One Question, you mapped only the models folder, why ? Is it problematic to map the others one at A1111, like the lora and uspscale folder too ? Thx
still waiting for the photoshop like UI for A111 XD its nearrr i feel it!
I use a variant, Fooocus-MRE
Does it work better with AMD cards? That would be awesome!
nothing works good with AMD cards ..... shame AMD ....
Thanks for another informative Vid! However I did make the suggested changes to the "user_path_config.txt" linking to the already downloaded SD checkpoints, but it seems to rewrite this file each time Fooocus is loaded with the run.bat file. Anyone else have this problem or fix?
I had the same problem.
check the path and pay attention to replacing \ with \\.
in my case this was what caused the rewrite.
I hope it will help you
Hi
Wonderful video
How did you manage to work in dark mode
- A1111 killer?
- No
No, Not a killer. But runs REALLY fast and has enough features for most basic stuff.
Yeah that was my reaction as well, there is nothing even close to A1111.
@@speaktruthtopower3222 The first startup of WebUI even on my old i7 7700 released in 2017 and HDD (not SSD) takes 12 seconds, right now I measured this: from launching the startup .bat file to automatically opening a new tab with WebUI in the browser. Subsequent image generations will also be comparable in time.
Obviously, a simpler and less functional tool without a couple of dozen plugins will be faster overall. It's silly to argue that.
But if you use all the features of the A1111, you'll use it for both basic and complex tasks. There is no need to install other software in this case.
I will say if you do have Windows 11 Feature Experience Pack 1000.22674.1000.0 or higher, 7zip and winrar files can be opened natively with explorer
Thanks for the great find! I just downloaded and unpacked and don't see a user_path_config.txt file in there. Any ideas?
same!
Thanks, but unfortunately generating is way too slow, I think its faster on my computer :) Any way to make this faster?
hmmm... Ich glaube fast die Pfade sehen so seltsam aus mit den doppelten \ weil die Verarbeitung intern sonst nicht funktioniert. \ wird in der Programmierung oft genutzt um Sonderzeichen die sonst andere Funktionen haben nutzen zu können.
Nicht das es irgendwas daran ändern würde das sie vorhanden sind.
When I click generate, getting the message --> "Waiting for task to start ..." . How to resolve? Please help
It's slow compare to automatic1111. I have 3060 12 gb, it takes 4+ mins to generate 1 image. I have installed couple time, changed os but still same issue.
I have a 3060 too and it takes 30s per image. Works like a charm. Faster than A1111.
@@AnandaFury hmm I have same card but it takes long time. Installed multiple times & changes OS as well. Nothing helps.
for ma also is fast @@CloudVideoYT
@@CloudVideoYT it is strange. I use the newest nvidia studio driver. Game ready drivers are worse. Win10 pro. Intel i7, 32GB RAM.
@@AnandaFury wait are you using windows 10 ? Well this will be the reason. I have windows 11 & i7 processor. But only have 8 gb ram
Can you do a video on loading more models to this?
just put them to models folder?
For me, generating stays at "Moving model to GPU ..." phase, and nothing happens anymore, even after couple of minutes. Someone knows how to fix that?
It's impossible to use, I just installed it and it consumes my 16GB of RAM! I have never seen in my life a program devour all my RAM. The GPU was what should matter, but apparently it just uses all the RAM without doing anything, just opening the generated link. I can't have 40GB of RAM or more, 16GB is enough for everything here, it's never been a problem. I have a good video card and I can't use it because it sucks up all the RAM, I don't know what else to do.
I am running win 10 with gtx 745 gpu and 12 gigs of ram and I have tried and tried to get this to work with no luck . The interface shows up and it renders one picture and when it hits 30 the picture closes and the program hangs, Iam tired of it. :O)
Why did he use parentheses in the prompt?
hi dear dude... Can we use these image in youtube videos. What is licence agreements?
I tried it, i can see using it if you had a laptop or low end pc. otherwise its not very good.
Who would come up with a way to reduce the requirements for the model =)
Half the world with 6-8gb vram
I'm getting an error at 'Moving model to GPU ...' 😥😥
Needs more ControlNet
Does anyone know why input image doesn't do anything for me, and how can i fix that? Did anyone have a similar problem?
👍👍👍
hmmm mine is white, there a settings tab anywhere to set it to dark mode ?
Hello. Thanks for tge instructions. I downloaded, but by running it doesn't generate anything, error , error. And, i not found user_path_config file in folder. And one more, my laptop is on Radeon graphics, is it important? Pls help
nvidia still has driver issues and generating is very slow - had to revert to april drivers and work ok again. I wonder if this is also the issue with Foocus
always use the studio driver and not the game ready driver!
Maybe a compatibility issue. Have you tried to re-install auto1111 with torch2?
Hey! When i change the path, is starting to download SDXL again.😒
I don't see the user_path_config file, do i need to create the file myself or maybe I need to run the bat file for the first time? edit: ah ok yeah i just had to run it for the first time, then it created the file for me.
did it have to download the xd model before it gives you the user_path_config.txt..im downloading the xd checkpoint it now and still no user_path_config.txt
@@hairy7653not sure. I already had the SDXL base and refiner models since I use them in A1111 so I just copied them over temporarily to Fooocus folder so it wouldn't have to download them, then i ran the bat, verified Fooocus was working, then i was able to edit the user config.
This seem like just a1111 with custom styles
It's a nice UI overall, but the fact that it loads the models EACH TIME you make a generation make it really really slow, i can make dozen of generations on A1111 between 2 generations on Fooocus...
I assume you are using SSD nvme like every nowadays person so load 6 GB of data to ram or Vram takes 1 second ...
..and like allways setup is changed it did not work. needed do magic that not install instructions have and you know if you once have done python venv
Where do I find inpainting models for SDXL, I haven't come across any in civitai
I just have 12gb Ram and 128mb Intel Graphics Card. Can I run Fooocus on my pc?
HI sir , I Have Almost 12 GB Of Ram Installed With Nvidia GTX 1650 But My Laptop Got Shutdown unexpectadley While Generating The Model. And Generating The Model Takes Too Long. Any Help would be Really Appreciable ?
running with GT1030 and 4gb VRAM
how do you open the UI onc you're done install
I've got an old Surface Pro (2017). It only has 128mb of VRAM but 8gb of RAM. Will Fooocus work on my system?
Anyone figure out how to change the port it runs on? I already have an instance for my discord that uses 7860.
Hey, how do you install refiners? There is an update Juggernaut XL V8 but i cant find the directory to place it in. Cant find it online either
i dont have the user path config file
Mine doesn't have the user option to add the location of the models as I'm using comfy ui too,
And when I run it has network errors, so is there a fix for it ?
when i click on download for windows, it sends me on 404 error page ! ='(
any hint ?
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? :)
Fooocus-RE?
How do you make changes to the generated images though?
hi my theme is white, you know how to change to dark?