Advanced Stable Diffusion Features in Fooocus
Vložit
- čas přidán 29. 09. 2023
- Learn how to do Image To Image (Image2Image), In Painting & Outpainting in Stable Diffusion using Fooocus.
You will need Fooocus installed and detailed instructions are provided here: • Create ANY Image with ...
Requirements for Fooocus: Windows & an NVidia graphics card with at least 4 Gigs of VRAM
Another youtuber showed me how to download Fooocus and get started and your two videos gave me all information I needed to be confident and start really using it! Thank you a lot, I hope you will have more in the future :-)
I'm jealous of the animated desktop background. Didn't know that was possible.
That is Wallpaper Engine. It is available on Steam for about $3 and has thousands of user created wallpapers, all for free - highly recommend
Fantastic
Wonderful stuff! Can you please tell me how enabled dark mode for this UI? Thank you
I am not on that computer right now to see exactly but the UI is just a web browser (Chrome) so that is likely just a Chrome theme or it is set to mirror theme the setting of the PC. I like dark mode so I usually set everything i can to it right up front.
@@holidayeffects3933 Thank you, Yes I think I can do that within Chrome
Ive been trying to find any documentation about how to control the number of steps in Fooocus and cannot seem to figure it out. Im prolly overlooking a simple setting but would appreciate any help from yourself or the community in figuring this out! I used to used CMDR and the UI is similar but different enough im relearning things
Under "Advanced" and the Settings tab, you will find a "Performance" section. That changes the steps but only within 3 possible settings: Speed is 30, Quality is 60 and Extreme Speed is 8
@@holidayeffects3933 ohhhhh, dang I was hoping to have more fine control.
I’ll have to try some other UIs then but for now it’s pretty solid. Thanks for the tutorials and answer!
Yeah Fooocus is all about limiting complexity but that means options too :(. The next closest is probably Automatic 1111. Not as simple but much more control. Comfy is super powerful but you need to build a workflow just to get an image out of it.
@@holidayeffects3933 I really want to try comfy. It reminds me of a few other tools I’ve used like Origami (FB prototyping tool, and Max Live 7 for audio visual stuff)
hey what hardware do you have that the model runs so fast? I can't run it on my laptop at all
It is a laptop 3080. I find cards like the 10 series really struggle (don't have a 20 series) and AMD cards & other platforms either aren't supported or have poor performance. The only exception is AMD ROCm support on Linux. This is changing but, for example, AMD just release ROCm late summer trying to catch up on this.
Anytime i try and use this i get the error:
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:1 and cuda:0!
It sounds like it has detected two graphics cards and doesn't know which one to use. If you do have dual cards, try disabling one and in any case i would update your NVidia drives to the latest. There is also a command line option "--gpu-device-id DEVICE_ID]" where device ID is probably either a zero or one (0 or 1) to specify the card to use
How do you use embedding with Fooocus?
Fooocus is designed to be a simple interface to Stable Diffusion that automagically gives good results with little user prompt engineering. For advanced stuff you need to use Comfy or Automatic1111
Hi , i have 8gb ram amd ryzen 5 . How do i make this software run on my laptop? Any changes i can do to make it run ?
Unfortunately Fooocus currently doesn't support AMD GPUs/APUs. One of the other interfaces, Automatic 1111 does have AMD instructions and there is a fork on Github (lshqqytiger/stable-diffusion-webui-directml) which I know works on AMD APUs. In fact, I actually generated images on a Steam Deck running windows using that. It was only using Stable Diffusion SD and was slow but it worked. For a limited but free site that will let you make a few pictures a day, checkout Leonardo dot ai (and there are many others, but Leonardo produces good looking stuff and has a limited but free plan)
@@holidayeffects3933 thanks mate
@@holidayeffects3933 They might have meant Nvidia card but AMD processor. I myself use an Nvidia 1070Ti with a Ryzen 7 and I am able to work with Fooocus. 2 month old post though, so I'm sure they figured it out or moved on haha
By buying a better one.
how to made dark theme
It runs in a web browser (I am using chrome) just set a theme in the web browser of your choice and you are good to go
When you say advanced, I thought you meant the dev debug mode in the advanced tab, these are just the regular features
Yeah it is confusing. Fooocus is so geared for ease of use that many common features are under "Advanced" or "Input Image". All part of their "keep it simple" philosophy
Is there a reason why you cannot use an AMD gpu?
AMD was just very slow in releasing their AI support on GPUs (RocM) and so it is rarely supported. However that is changing and Mac M1 and AMD support is on the roadmap for Stable Diffusion and Fooocus. You can get Stable Diffusion to work with AMD GPUs by using the “stable diffusion webui directml” project on GitHub. That is what I used to run it on a Steam Deck (link in the description) czcams.com/users/shortsyS2CvFeOnco?si=17vbJjURxCwAmcEy
Nod-AI's SHARK is super fast on AMD and runs on Windows. For the time being, it can't do tiled images though. Definitely worth the try though.
@@moravianlion3108 it looks like AMD is putting money into Ai.
Always nice having multiple programs to choose from but I realy don't know whats the point of fooocus. At the moment it's a heavily watered down a1111 with styles to chose from per default. Am I missing something or is thats "it" at the moment?
Like Midjourney it does automatic prompt expansion so prompt engineering isn't as necessary and, as you mentioned, it has a bunch of built in clickable styles so it is newb friendly. Basically it makes it easy to get good results for people who are new to Stable Diffusion. No huge draw for A1111 or Comfy pros, it is targeting a different audience
Couldn't you make a video teaching how to download an app for free that turns 30fps into 60fps?
wheres the image to image process??? vis a vis Stable Diffusion....
It is in the newest Fooocus. Just make sure you are updated, click on Image Input box and go to the Image Input. If you click on "Advanced" under the 4 image slots you will be presented with several options in addition to the base Image to Image features