Easily Create Custom Styles and Wildcards in Fooocus

Рет қаралды 7,223

Күн бұрын

Enhance your Fooocus with new ways of using Wildcards and customizing your own Styles. Easily streamline your stable diffusion workflow by creating styles from your popular positive and negative prompts and creating wildcard trigger words out of your character descriptions.
Fooocus is the easiest Stable Diffusion SDXL Image Generator out there. If you have yet to give it a try download it here: github.com/lllyasviel/Fooocus
-----
00:00 Intro
00:18 Custom Styles
03:18 Custom Wildcards
07:01 Inspiration & Ideas
-----
Check Out my Full Fooocus Tutorial Playlist!: • Fooocus!

Пікірлер: 43

@JumpIntoAI 3 ай бұрын

Something I forgot to add was a new set of style examples created by a user named Irmagon. Check those out here: docs.google.com/spreadsheets/d/10UXU59fBA5wZrcEEvgv1MTf1MSNTCB5PT4fNxsAst8E

@Zonca2 3 ай бұрын

Is there an easy way to import this into fooocus? Other than copypasting everything yourself? could you maybe share the json file 😅

@JumpIntoAI 3 ай бұрын

@@Zonca2 lol, well the idea is once you create them they are in there to use forever. The stuff i showed was just quick examples. I don't think there is an easier way to make them that I know of. 😥

@hmm-m 3 ай бұрын

@@Zonca2 Those are the styles that are already in Fooocus, so what's the problem? It's just a gallery for easy reference. I'm only not sure about 2nd tab but I understand it's the same with only sampler changed to LCM.

@dguillemette 3 ай бұрын

I started with ComfyUI last summer, but switched to Fooocus few months later. One of the only things I missed was the wildcards I used in ComfyUI (e.g. "{blonde|light brown|light dark|dark} hair"). Fooocus wildcards are an even better format for my needs! Thanks a lot!

@b.s.design 3 ай бұрын

I didn't even think about the fact that you can create your own universal styles. Thank you so much for this tutorial!

@JumpIntoAI 3 ай бұрын

Yep! Glad to help.

@diarioia 3 ай бұрын

Last week I bought a new PC (16GB VRAM), downloaded SDXL and watched every single videos in your channel. I love learning with you. Excellent videos. Thanks.

@JumpIntoAI 3 ай бұрын

You are very welcome!

@daviddragonsage4874 2 ай бұрын

Thanks for a very inspiring video! I have done a fair bit of work with wildcards but did not realize they could be nested. Also, the idea of using one entry wildcards for favourite positive or negative prompts is brilliant!

@JumpIntoAI 2 ай бұрын

Glad to help!😊👍

@tomasholmgren9655 3 ай бұрын

Watched all your tuts and they are awesome, calm, i formative packed with awesome info. Keep em coming and thx

@DarioCapizzi 3 ай бұрын

Era proprio quello che stavo cercando e la community non era riuscita a rispondere. Grazie!!

@gebuildersdeveloperspvtltd3846 3 ай бұрын

Again simple, useful and to the point 👍

@mariopadilla1445 3 ай бұрын

Thank you for this! I’m gonna make specific “styles” to pair with some loras to streamline my process.

@Vahidk2222 Ай бұрын

Thanks for all this useful tips, all of them are life saving! 🙏🏻👏🏻

@PETERFILM 3 ай бұрын

Your channel is a godsend! Thank you very much for the video!)

@JumpIntoAI 3 ай бұрын

You are very welcome!

@TomiTom1234 3 ай бұрын

Again awesome tutorial. I am now using Fooocus instead of A1111 after understanding it better, thanks to you! The Inpainting alone is easier and more effective in Fooocus , not to mention the speed for even low vRAMs.

@JumpIntoAI 3 ай бұрын

Agreed, if nothing else the Inpaint sells itself.

@hmm-m 3 ай бұрын

@@JumpIntoAI Maybe after some more vids on Fooocus, we can move together to more difficult tools like exactly A111 or after ComfyUI? :) So far, I limit myself to Fooocus, this is where I start my adventure with AI but planning to use more advanced tools like the mentioned ones in the future.

@AlexLepeshkin 2 ай бұрын

In some styles (in the Prompt) there are words enclosed in parentheses. For example: "graphic, (blur, blurry, bokeh:1.3), text". What are these words in parentheses for? Which means ":1.3"

@JumpIntoAI 2 ай бұрын

words in parentheses means you want to add "weight" to those words. Meaning you want those words to be more important than the words not in parentheses. the numbers are how much more weight to be added. usually using 1.0-1.5 is common.

@LunaNovaChannel Ай бұрын

Can we tell it what to use from inside the list?

@canelj 3 ай бұрын

This is exactly what I have also done during couple of weeks. BTW, do you know is there any method to loop all, or random, predefined styles like what we doing with these prompt wildcards? I would like to see what Jess look a like in random fooocus styles. To get inspirations etc.

@JumpIntoAI 3 ай бұрын

I don't know of a way to get fooocus to select random styles. but using the wildcard idea (with some work) copy the prompts out of the style files and create a wildcards out of that.

@hmm-m 3 ай бұрын

Would be also awesome if it was possible to queue the jobs. Like, not only "sample 10x" the given input, but after that take care of my 2nd input with different parameters, then 3rd etc. You could leave the PC to bake some different prompts for hours with no need to come back and click the next set.

@hmm-m 2 ай бұрын

By coincidence I discovered how to do it ^^

@Denko-si4tz 2 ай бұрын

Appreciate the videos! Any idea if there's a way to do something similar to wildcards but with a json file? Like __jsonFileName: propertyName__? I know this is essentially what a custom style would do, but as you add more and more that list keeps getting more absurd... Speaking of which, is there a way to remap the styles location? neither "path_styles" nor "path_sdxl_styles" seem to work :|

@JumpIntoAI 2 ай бұрын

Right now no, but those features could always be requested on their official github page. How easy/difficult to implement I couldn't say. The styles list was fun for me at first but I've removed many of them on the fooocus I use (not the one i use for recording). to make way for my personal ones.

@Denko-si4tz 2 ай бұрын

@@JumpIntoAI That's kind of why I wanted to change the path, I'm not sure if it happens with every update or what, but I've deleted the damn styles too many times already and they keep coming back...

@SushilKumar-fb2mw 3 ай бұрын

Thank you so much! :) can you please make a video how can we copy facial expression? Thanks

@JumpIntoAI 3 ай бұрын

On the list! 👍😁

@AlexLepeshkin 2 ай бұрын

What does 'RAW photo' mean in your example ("prompt": "RAW photo, {prompt}, 8k uhd, dslr,)? Where can I read about it?

@JumpIntoAI 2 ай бұрын

a RAW photo itself is a type of format used in high level photography/filmmaking. It's just another word to prompt for better pictures like using 8k, dslr, HDR, etc. Or putting the name of an expensive camera. That prompt i used was the suggested base prompt for the popular model Realistic vision. How "necessary" some of those are is debatable. Especially in SDXL.

@wittchow 3 ай бұрын

how are you using the realistic vision model when it is SD 1.5?

@JumpIntoAI 3 ай бұрын

I was only using the suggested positive and negative prompts as an example, wasn't using the model. Although it can be used as a refiner.

@kingdompig1 2 ай бұрын

Is there other sites like Civitai that might have Loras that weren’t posted in Civitai?

@JumpIntoAI 2 ай бұрын

huggingface has models that civitai doesnt have. but it is not very user friendly so they can be hard to find.

@StoryFBMizostory 3 ай бұрын

I just got gpu, and i wanted to try this ia image genaration for novel, cover photos mostly for romance novel. with your guide i made some good photos but i need asian face, it wont gave me good result, most of them ugly, any tips or model or promp, can you give me. Thank you

@JumpIntoAI 3 ай бұрын

HelloWorld XL checkpoint is good for Asian faces, just make sure you are prompting for specifics, age, ethnicity, even specific country/province. as well as saying beautiful/attractive, etc. If the image comes out good but the faces look a bit undetailed (Usually images where the face is not close to the camera) this is where using inpaint and the Improve Detail (face, hand, eyes, etc.) setting comes in handy. Mask the face (one at a time) and select the "Improve detail" setting and type the kind of face you want in the new "inpaint additional prompt" text box. do that for each face in the image. make sure only 1 is masked at a time.

@StoryFBMizostory 3 ай бұрын

@@JumpIntoAI thank you very much, i have problem with python. Its using around 90% of ram, i am using 16gb ram, but my vram usage is only about 20% most of the time. I google it some other have thia problem too but no solution, if you have solution please help, thanks again