Thanks for this, just installed Fooocus but now want to try the MRE version with controlnet, good stuff. FYI instead of copying over your existing models, if you run the run-mre bat file one time and close it before it downloads the models, it will create a user_path_config txt file that you can edit and change the directory to your existing models you used in other version of Fooocus or A1111 etc. Just make sure to follow the format using the \\ between each folder.
Hi everyone! Hope you enjoy this Fooocus-MRE version. I do want to make it clear that this version is a "Fork" of the original Fooocus so things may not be exactly the same as the original. Use either one and don't confuse the 2. This version was meant for those of us that wanted a few more advanced features. If you want something easy and simple midjourney style stick with the original Fooocus that can be found here. github.com/lllyasviel/Fooocus. I think it's amazing to have such an amazing open source community giving us so many options!
LOL, been using it for a while now, a few days. Not this version, for which I'm grateful that you've brought our attention, but standard Fooocus. It is awesome. I'm currently extracting the zip right now. Thank ya for the video and information!
Nice then you’ll feel right at home while enjoying the added features! Honestly I just happen to stumble across it this past weekend so I did a video on it. Been using the original version too 😊 let me know your thoughts once you’ve given it a try!
Short of the load time of a model *just like every other,* this thing goes just as fast as you said. It has the ctrt+enter which is incredibly nice. Also, being able to type in something like "photo" or "horror" and filter all those styles is a big plus. Having the AR listed like 3:4 as opposed to 1234x4321 only is great too. It's nice if you know you want a 4:5 portrait for example, Then the metadata being able to be saved in the image or in a file in itself is awesome. Thought it might work with Civit, but it has yet to pick it up. Still cool to have, though. I haven't messed with the img2img and CN stuff yet, but I plan to tinker with that later. @@MonzonMedia
Great feedback! I definitely share the same thoughts. I think in time it will only get better and I'm happy to not have to wrestle with settings and optimizing things just to get things to run properly! Going to do a more in-depth video soon! Appreciate you bud and thanks for the support!
I don't know which part was the best, the video itself or you every time you said Fooocus 😅. Thanks for showing us another good program to do SDXL quickly with some decent options! Loved it!
😁 sometimes I amuse myself a little too much! Gotta have fun right? So apparently this uses a back end similar to comfyui that's why it's optimized really well and the original is the guy who created the original controlnet. I love its simplicity and I'm hoping for more "simple" features to come soon!
Yes you have to have fun! Absolutely! Nice to see it's core is based on Comfy probably why it's so speedy! I'm sure he'll add more bit by bit, thanks again for the look!
Always great to see updates and additions to any platform, I really like the simplicity of fooocus, it's a nice change since I cover the other platforms too.
Another Gem 👍 Another late night 🍷 Thanks, this is going to fill the gaps in Comfy which I love! But Styles and IMG2IMG is what I was missing, cheers 🍺
What you mean these days, it's been a rush of development and new things for months now, you can't keep up, I have a list of todos as long as my arm lol
My excitement was short lived, although the install went smoothly, presenting me with a ready screen to enter a prompt, the rendering took an inordinate time, in fact would freeze. Comfy UI works like a charm, have an RTX 3070 8 gig GPY with 64 g RAM. So go figure?@@MonzonMedia
hhhmm intersting, I've tried both the original fooocus and the MRE version and it's almost as quick as comfyui for me and it's said to have the same or similar backend. I know there was an update yesterday perhaps something changed? Perhaps try a fresh install? Did you try it before going into the advanced settings? Or perhaps something is off in the settings?
You look like a good guy who has created an excellent video. I would like to ask you for an explanation. Last night, I installed the basic version of Fooocus. I'm an architect by profession, and my idea was to import a render from Blender for the house I'm working on and use Fooocus to get different designs for the same object. For example, if I have a picture of a house with a brick facade, I want to see how it would look with a wooden facade or what it would be like if I replaced the aluminum windows with wooden ones, and so on. The basic version of Fooocus allows you to input your own image, but the manipulation options are quite limited, and the results I obtained were rather weak. I see that the Fooocus MRE version has an Image 2 Image option, so I'm interested in how useful this could be for my needs. Is it possible to get precise results with any version of SDXL, in the sense that it doesn't completely change the appearance of my house but only specific details? I'm not sure if it even makes sense for me to use these tools if the essence is to generate an image based solely on text, and if the given reference image always undergoes significant changes compared to the original version. Thank you in advance if you have any advice for me.
i like more standard version fooocus it has image prompt tab better it can generate mixed of 3 images for example subject from image 1 , faceswap from second image , background from 3rd image but in MRE version that is not posible i guess bcz there is no input field for faceswap or bodypose normal version is better as features mre is better ui i choose more features so i go with standard
Yeah I’ve been covering the original focus lately mainly because the developer of the MRE version stopped updating it. Which is unfortunate because it was a good alternative.
With introducing the AI to the masses and becoming accessible it was presented as something that will never be capable of creating art. It is somehow scary that the first thing it started to create was a digital art...
I think it's in our nature to create so since we created AI.....makes sense. I can see how it can seem scary but I think it's just because it's all so new still.
*UPDATE* Unfortunately I just found out the other day that this version will no longer be supported. I did a video on the original Fooocus and mention it there. However the developer of this version is hoping to work with the original Fooocus developed to see if any of the MRE version could be implemented. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-w5EA358IPCs.htmlsi=tVSviqAvLGfI7WST
Merci, pour une raison qui m’échappe ,j'ai des erreur de compilation lors des instalation des AI, donc je l’orient vers les solutions Colab,(souvent les biblioteque que ne veulent pas ce mettre a jours.). Thank you, for a reason that feels me, I have errors in compiling the 'A's 'Anail', so I'm orienting it towards Colab solutions, (often the libraries that are not wanting what is going on.)
@sergiykravchuk Quality settings and resolution can both be set to maximum for the given checkpoint. I normally do 60 steps for quality on a model that isn't a Turbo version . More steps than that don't make much difference. You can go to 1280x1280 on most SDXL models without issues for the resolution.
@@isaacnaughton5206 thanks, one more question i have a problem with fingers and human faces, how to solve this problem, I also have a problem with buildings, cities (far) and cars (the cars sometimes looks not realistic, can I do something with it, maybe use another Lora, I use as a default and "realisticstockphoto"
You want to have it checked on for that look. See the statue example here github.com/lllyasviel/Fooocus/discussions/117
Год назад
@@MonzonMedia It's confusing that lllyasviel first said that turning off "Raw" mode - i.e. the default state brings "Midjourney-style", and then it's now included in "Style" under a new name as "FooocusV2". However, now here in the MRE is "Prompt Expansion" to make this very pleasantly simple program not so simple after all...
That's a recent change for the original fooocus literally just yesterday, remember this is the MRE version so it's not exactly the same. It's a fork of the original. You can always make suggestions to Moonride github.com/MoonRide303/Fooocus-MRE/discussions If you want simple I'd stick with the original version.
One of the nicest guys on RU-vid and in the stable diffusion AI community. Really enjoy watching your videos. Was looking for the image to image version Fooocus. Ta! Now have three versions so will start rendering and see how it all goes workflow and process wise.
Very kind of you to say! I really appreciate the support. Just note for this version the developer is no longer updating this version, I believe he's still maintaining it but no future updates to come. That's why I haven't been covering. However I've been covering the main version and there are 2 other forks I will be introducing to the community. Sounds like you are aware of them already.
Thanks for the video, very interesting. I don 't understand image generation and writing scripts at all . But somewhere I looked that you can take a picture insert and render pictures on this topic and already choose from them what you need. Please tell me in the Fooocus program is it possible to do this ?
In what context? Flying? Posing in front of the plane? Honestly, it's just describing what you want to see in it's simplest form, then adding some key words as modifiers.
I'm waiting on the Fooocus team to officially announce AMD Windows support, which is still in the "coming soon" stage.,I tried the original version, but it full loads my AMD RX6950XT 16g vram with 95 ° just gen for 2 images...
Oh that card will do you nice! For standard generation, really nothing to worry about. The added 24gb vram on the 4090 will come in handy once you get into using advanced tools. Currently I'm on a lowly 3060Ti, good enough for most things but not powerful enough to train models.
I've been using the original for a week now and created some nice images but as my pc graphics has just below minimum specs it was very slow and had the odd crash.. so until I can afford a better graphics card.. how do I go about uninstalling this.. is it ok to just delete the files or is there a particular process for removing .bat type files, scripts, software.
Nah not really....for simple image generation it's a decent alternative but unfortunately I just found out this version is no longer going to be supported. I did a video on the original Fooocus here and mentioned it. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-w5EA358IPCs.htmlsi=mNOO6Am317VMZJMG A1111 is still king, since it's just more versatile than anything else.
@@MonzonMedia Hi, I watched your new video and installed Fooocus. But I have an i5 with 8 GB RAM and a GTX 1060 w/ 6 GB VRAM. Not enough for that. Anyway, it was a good experience. Thank you!
Fooocus isn’t made to train models but you can use SDXL models with it. It doesn’t use SD1.5. If you want to be able to train you have to use Automatic1111 however you need an24gb vram graphics card.
Yes it is however I've only tested it with the default ControlNet it comes with. Will do more testing soon and there still isn't anything for open pose.
How much memory do you have? All stable diffusion platforms use a lot of ram, it’s no secret. Recommend is 16gb at the least but more likely 32gb. Also it will eat up gpu vram. This version is no longer being supported check out a recent video I did on the original fooocus.
@@MonzonMedia I've always used the SD XL and automatic 1111 without any problems for a long time! I have 16gb of ram!!! RTX 3050 8GB - 16GB RAM It's unusable! As soon as it opens it uses all the RAM! I opened a discussion on github but I still didn't get any answers and I saw someone also having the same problem. Everything very strange! I downloaded the latest version of Fooocus!
As I said could be because it’s no longer being supported. I’m going to create a video soon on a new fork Fooocus Control that has ControlNet. I really liked this MRE version but it’s too bad it won’t be updated. Generally speaking you should consider upgrading system ram regardless of what SD platform you will use. Fooocus uses a comfyui backend so fooocus should be decent with handling ram.