Ruined Fooocus with Flux.1 schnell (Part -1) - Generate high quality images using AI

[@Aininja94]

🌟Get Midjourney to generate your images: 🌟 www.aiarty.com/midjourney-prompts/?ttref=2409-aia-mj-ytb-AN-incl-wxr 🤯Aiarty Image Enhancer, the best AI tool to enhance and upscale images:🤯 www.aiarty.com/?ttref=2409-aia-mj-ytb-AN-incl-wxr

[@RuneW0lf]

Author of RuinedFoocus here! Absolutely awesome video man, thanks for the showcase!

[@Aininja94]

Thanks so much for the kind words! I'm glad you enjoyed the video, and I appreciate you taking the time to comment. It means a lot to have the project author's seal of approval!

[@SetFullmaisonze]

Please, .ipynb RuinedFoocus colab notebook

[@RAMATW2023]

I am confused, I followed his step by step but my models tab looks 100% different, This is a recent video so it does not make sense. Am i missing something?

[@_VloneX_]

Ruined Fooocus is awesome, I dont know how you made it follow prompts so well, its more clear, I dont know if its my imagination or what but its awesome. In the original Fooocus you could press I think right ctrl+up/down to change how much it focused on that word, it would be cool to get that back, but its better in every other way, thank you

[@nhldesktop]

How to use the unet, clip and vae version? I downloaded them all and put in the right folders but can't find the in the ui

[@AbdulKhamis]

Thank you very much for this tutorial! I was waiting inpatient for Fooocus to adapt Flux officially (If they will someday!). This Ruined Foooocus is what I needed!

[@Aininja94]

Actually, getting in added into OG fooocus is quite a task. R-Fooocus is using confyUI inside to get flux working. Still complex though.

[@tincheinvarte]

Thank you very much I did everything as you said, and I was able to install and start the program, it working as yours, I used the exactly same settings as you, except it works much much slower, I have Nvidia GeForce RTX 3070, AMD Ryzen 9 3950X 16-Core Processor 3.50 GHz and Installed Ram 64GB. For people who has problem unpacking it really requires winrar it unpacked like a charm, but with winzip it shows error.

[@Aininja94]

Well you got bigger RAM and processor. I am using 3700k 😅

[@tincheinvarte]

@@Aininja94 which graphic card you using??? The GeForce RTX 3070 has 8GB RAM

[@Aininja94]

I am using 4060ti 16gb. 3070 is more powerful but it gets limited due to low vram.

[@tincheinvarte]

@@Aininja94 Thanks for the info

[@videosfeoscomotucara9038]

You are a genius thanks bro, I will wait for your version in colab since I don't have a graphic card

[@Aininja94]

I will be doing in colab, the thing is, it has limited resources for free user. Earning from KZbin is still far for me. Your support is appreciated, I will be using KZbin money for colab trial and error to get the best for other users.

[@souravroy8834]

@@Aininja94 Can you do for kaggle? colab kind of very limited .

[@SanchezGodsent]

Hello! I need a help. When i generate in ruined fooocus via flux (s or d) , i got a blurry picture. How to solve it?

[@Aininja94]

The custom performance setting has to be done. It's shown in the video at 4:31

[@SanchezGodsent]

@@Aininja94 nice thanks

[@JoKer-ds5jg]

THX, but i need help please "WARNING:root:lora key not loaded..." whats wrong

[@Aininja94]

Can you please post this on the Ruined Fooocus discussion on github. Will be able to check the error if you upload a screen shot. Or can you try a different lora? The application is not able to get the meta data like lora key.

[@JoKer-ds5jg]

@@Aininja94 Thanks for the quick response and help. I installed everything as described. At first I didn't have a config, but I downloaded it, deleted the standard Lora and then it only worked without Lora. I deleted everything again but will try again

[@vic16161]

Any way to run ruined fooocus on colab?

[@Aininja94]

Working on a video, flux on colab. Should be out after 4 days

[@christoherright6430]

Can I use it with AMD card?

[@Aininja94]

I read some post on Reddit, it didn't worked on amd cards. I am testing a application right now, it should work. I will make a video in few days.

[@dqschannel]

What app to you use to give detail on CPU, Disk, GPU numbers?

[@Aininja94]

I am not using any other application then what windows offer 'task manager'. Nvidia does gives a better application just to monitor performance.

[@dqschannel]

@@Aininja94 I have an Nvidia card but am not aware of such an app. Is that something I can download?

[@Aininja94]

They have a application 'nvidia geforce experience' there is a option in game overlay, switch it on. Then go to settling then go to hud layout. Then performance. It will ask you where to show your computer hardware status. I used it while playing game, but it should work in this AI things as well

[@gerva204]

Thank you very much for video. but a question... how do I change the VAE?

[@_VloneX_]

I was able to use inpainting without the Vae with Ruined Fooocus

[@gerva204]

@@_VloneX_ yes, i also tried it on ruined, but, almost at the end of the generation, (when it loads the vae) it gives me a distorted image. i think it loads the vae of sdxl.

[@Aininja94]

I am not sure, the vae is in models/checkpoint folder. The file can be changed and renamed.

[@gerva204]

@@Aininja94 ok, thank you

[@_VloneX_]

@@gerva204 I use the default one on pony checkpoints and it works when it used to not on original fooocus, and when pony is meant to not be compatible. I hope you figure it out because that was one of the plusses for me on this new Fooocus

[@vitaminprotein334]

Can you please da a video to install fooocus on window amd

[@Aininja94]

Ok, will add in my fooocus playlist in couple of days.

[@Aininja94]

I tried getting amd card, but failed to get one. I am still looking for card from someone for a video. I made a colab video, upload just now. May help you getting the application running

[@MillennialKiwiGamer]

sadly doenst install correctly through the official stability user interface matrix, must be some kind of error recently

[@Aininja94]

May not be in matrix, this is the second time i am reading this. How about getting the file download from GitHub as shown in video, but i guess you have to download the checkpoint again.

[@MillennialKiwiGamer]

@@Aininja94 I run multiple uis through the stability matrix I will wait for an update if this is a common issue

[@dqschannel]

Is there a manual as this is very different than OG Fooocus?

[@Aininja94]

I did not found any maintained documents, this is the best. github.com/runew0lf/RuinedFooocus/wiki/PowerUp

[@dqschannel]

@@Aininja94 will check it out. Thanks

[@Dilfin90]

Guys, who knows why it doesn't start without internet? I have very bad internet, there are frequent outages, and at such times, I just can't run the program. It crashes on startup without an internet connection.

[@thiagolobo11]

Where did you download the Flux models from? It would be nice to post the link and tell exactly how to "install" it. Just throw it in the Checkpoint folder and that's it?

[@Aininja94]

Oh, didn't thought of, editing the description - adding the links. Yes, you need to download the file and keep it into checkpoint folder. Maybe you have to use the refresh button below model section to load the model if not happened automatically.

[@ronbere]

@@Aininja94 if i have this checkpoint on another HDD, dont have any config file for path settings?

[@Giorgio_Venturini]

@@Aininja94 I join in the request for the link for the Flux files. Thank you

[@Aininja94]

I tried doing the same in comfyUI it's didn't worked. I don't think it would in fooocus it will. There is a think called symbolic link, that will work.

[@ronbere]

forget it ...RuinedFooocus\settings\paths

[@rennynolaya2457]

Is there a way to use inpaint and control net? with flux?

[@Aininja94]

Not there yet, I uploaded a video on it just now.

[@Muz889]

​@@Aininja94Can't find the video. Link?

[@Aininja94]

This is the second video, inpaint is not yet for flux. kzbin.info/www/bejne/mafGo6loZqZ7hZo

[@TMO2913]

does it support nf4 or gguf?

[@Aininja94]

I tried fp8 only, it works. Rest like fp16 gives errors trying to understand as its just experimental. gguf not there yet.

[@renzocheesman6844]

will this app ever be on Pinokio?

[@Aininja94]

Sorry man, no idea about it.

[@dqschannel]

I have downloaded several Flux models from Civitai and so far only flux1SFp16_v10 has worked on Ruined, all others give an error and it loads the default model. Have you run into this issue?

[@Aininja94]

I posted few hours ago about the working models in RFooocus. With an image, did you downloaded from that list?

[@dqschannel]

@@Aininja94 I have not seen it. Will look for it.

[@Aininja94]

kzbin.infoUgkxNHjNQqN5MQk5wDNpjGHF3N30TfpT4OKX?si=0kMLbWOX5itYreM7

[@Aininja94]

I wonder what sampler and schedular combination would work for this models. That's a question remains.

[@dqschannel]

@@Aininja94 guess trail and error. One of these suckers is over 20G

[@sxdev]

If I follow this it just creates like a pattern, its doing something but not a picture

[@Aininja94]

Your settings may not be correct. Did selected custom in performance shown at 4:31

[@RudmerCR]

No windows AMD support?

[@Aininja94]

Few have tried but didn't succeed, based on Reddit post. But SDXL worked, you can check on Reddit.

[@dqschannel]

Hows does this work on a system with 32G RAM but only 4 VRAM?

[@Aininja94]

Flux will need atleast 6gb of vram, as per the user comments I saw on Reddit

[@dqschannel]

@@Aininja94 I used it and it took over 25 minutes using just 16 steps.

[@Aininja94]

Bro you got serious patience and you made it work. Did you tried the schnell? It even works in 2-4 steps.

[@dqschannel]

@@Aininja94 trust me not easy waiting that long as I am used to faster time with OG Foocus. I am trying it out now using 4 & 8 steps using flux1SFp16_v10. and a flux Lora. using 8 it cranked out an awesome image in 12 minutes. So now trying same with 4 steps.

[@dqschannel]

@@Aininja94 downloaded the Schnell which that sucker was over 20G and the oupt sucked ass. Even at 25 steps.

[@AntonioMaciaMartinez]

Hi master! I want to inpaint but looks img2img dosnt work?

[@Aininja94]

That's the reason I am making a second video. The inpaint needs some work. And other feature needs some work

[@AntonioMaciaMartinez]

@@Aininja94 ok master I wait for it

[@Aininja94]

Sir, I had upload a video explaining what works in flux and what not. The inpaint feature does not work with flux model, there is some work to be done in ruined fooocus to make inpaint working with flux. It works on SDXL as per the development team.

[@AntonioMaciaMartinez]

@@Aininja94 thanks master!

[@lejimmycadet]

Super

[@DailySportsipl]

Download and Install complete but don't open. by default window.

[@Aininja94]

Please share the issue on git discussion section. Will try to figure out the issue.

[@stillsbyjann]

can I install it in Pinokio?

[@Aininja94]

I did not tryed Pinokio yet. I wilI will update once tried.

[@stillsbyjann]

@@Aininja94 checked it, didn't find it

[@StellarNear]

It would be Nice to have this in Pinokio so we can replace old Fooocus

[@sephtube]

No mac installer?

[@Aininja94]

It can be done by using git.

[@dongytuetinh]

please give me the link colab fooocus. thanks bro

[@Aininja94]

I don't think it's there. I will be making a collab installation in future. Just struggling right now into python code making application and stuff.

[@Nikida18]

RuinedFoiocus vs Fooocus. What's the difference?

[@Aininja94]

Flux support in RF Enhance feature of Fooocus 2.5 This is the major difference. There is second part of video I am working on. RF has many others feature which I have not used yet. Some are only for SDXL. Some can be made for both SDXL and flux. There is lots of work to do.

[@Nikida18]

@@Aininja94 ok thanks. I will wait for the new video. You have a new subscriber 😬

[@Aininja94]

Awsome, I am just working on final thought after using flux. Video should be out in few hours.

[@johnny5805]

Did you read the comments on the Flux model page on Civit AI ? It takes MINUTES to generate at 1024p image using a 4090 !!!!! And requires around 20gig of VRAM to work !!! And the model itself is 22 gig !!! It's no better than the realistic SDXL models, it's just better at the cartoony, or nerd stuff like you generated as an example. Definitely not worth it until they make it faster, smaller, and more optimized.

[@Aininja94]

In the video schnell took 25-30sec, In second video dev took 55-60 sec after first image. I believe in trying by myself on my 10 year old pc with 4060 rather then reading comments everywhere like a nerd. 😆