Stable Diffusion - FaceSwap and Consistent Character Tips - Part 2
ฝัง
- เผยแพร่เมื่อ 25 มิ.ย. 2024
- Master Fooocus and Stable Diffusion for Creative Image Generation! 🎨 In this Stable Diffusion video I go over some more tips on consistent character creation and how to use FaceSwap and get better results. Got some tips from comments and did further testing. #Fooocus #StableDiffusion #DigitalArt #ImageManipulation #CreativeTools
This is part of my ongoing series of videos on installing and using Fooocus for Stable Diffusion. Fooocus is a great interface for using Stable Diffusion and is very easy to get started. It also will run on systems with only 4GB VRAM.
I am running on an i5 32GB RAM Nvidia 3070 8GB VRAM
Face Grid Image: / face-grid-for-105881604
If you find the video useful and would like to tip, TH-cam does have the Super Thanks option. Those are greatly appreciated so I can have my coffee!!
☕ BUY ME A COFFEE: ko-fi.com/kleebztech
Also I am a verified creator with Brave and do accept BAT for users of the Brave browser.
Download Fooocus: github.com/lllyasviel/Fooocus
How to install Fooocus video: • Stable Diffusion - Dow...
Fooocus Beginners Guide: • Super Easy Stable Diff...
Fooocus Upscales and Variations: • Stable Diffusion - Foo...
Checkpoints and LoRAs the Basics: • Stable Diffusion - Che...
Fooocus Poses and FaceSwap - Image Prompts: • Stable Diffusion - Pos...
Zoom with Outpainting: • Stable Diffusion - Zoo...
Inpainting with Fooocus - Don't Regenerate, Fix!: • Stable Diffusion - Inp...
Background Replacement with Fooocus: • Stable Diffusion - Bac...
Fooocus Tips and Tricks (and AI Hands) • Overlooked Fooocus Tip...
Stable Diffusion - Styles in Fooocus - How do they work?: • Stable Diffusion - Sty...
FaceSwap and Consistent Character TipsPart 1: • Stable Diffusion - Fac...
00:00 Intro
00:20 Create Grid
03:25 Create Character
04:30 Upscale Grid
05:45 FaceSwap
09:55 FaceSwap and Inpainting - วิทยาศาสตร์และเทคโนโลยี
If you find the video useful and would like to tip you can buy me a cup of electricity for all these image generations. It is greatly appreciated! ko-fi.com/kleebztech
Very, very Cool! Thanks!
Incredible!
Most of the images I generated was closeup, can you make part 3 for different focal points and blurriness of the background.
This is a process I had never thought of before. It's absolutely fantastic. My long-standing concerns have been resolved. You are the best.
Dude I love your videos!! Great job. Can you also do a video on consistent object/product. I know you try you can do some magic!
I have been thinking about it but not sure of a great way. Will keep thinking.
Can you share that face grid man? It makes things a lot easier for the whole lot of us. Else, us beginners will first be learning how to produce face grids!
Here you go: www.patreon.com/posts/face-grid-for-105881604
@@KLEEBZTECH Thank you buddy! 🙏❤️
great
Another tip: save an inverted copy of any good face that is at an angle. That way you have a left and a right facing version
Good idea. But I think Faceswap might be able to deal with that from my memory. Will have to test again. I don't use it that often.
Thank you so much for sharing tips about Fooocus!! What about consistency clothing, in different angles? Is it possible? Thank you!
I have some ideas for a video on that very soon. It is somewhat limited on what you can do but for some purposes it could be useful.
Very nice, will try this myself! Quick Q: at 14:03 when you increased the weight of the Big Smile prompt, was that video editing that made the brackets and value pop up, or is there a button in Fooocus for it?
3 minutes into this video. Just one correction to that video, when reselecting you don't have to highlight the text again you only have to click somewhere within the text. th-cam.com/video/QrkpFa4ArVM/w-d-xo.htmlsi=HT-GQT1idiQ9rmEf
@@KLEEBZTECHAh brilliant, cheers mate!
i just discover this tools it's amazing and thanks for your tutorial video it make easy to understand, is image generated by fooocus is free to use for commercial purpose?
Usually but that can depend on the model used. SDXL was released under licensing that allows for commercial use.
But I am not a lawyer. 😁
Hey, great video! Can you explian me a bit More of random? Should I use a big number for seed or small number? Thanks!
Really doesn't matter. Every seed is unique. But when testing things it is a good idea to use the same seed. If you use the same seed and all the same settings and models you will get the same image.
Fantastic video! How can I get consistency for body shape/size? For example, a larger/curvier woman?
Not always easy but one of the big factors will be using the same model/checkpoint. You could use a mix of celebrity names to combine two celebs that might be similar size but that is not always going to work. image prompts can help a little. Good description of body can help but you will always have some variations that you will have to deal with.
how do you get this, when you prompt an image, where the person looks directly to the camera, and then it should look away, to the side?
In addition to the table created at 4 different angles, is it possible to add expressions of laughing and angry at these angles? So that we have 8-10 models of the same character in total ?
It can be difficult to get that many. I am still experimenting and looking for better ways. The multiple expressions with the grid I always had some difficulty with.
I watch a lot of videos. Among the developers I watch, maybe you are the closest to achieving this. congratulations.@@KLEEBZTECH
If I already have the face I want in an image previously generated, and I wanna do the grid of 4 images. 3:27 I put the face with faceswap but the face is not exactly the same. Is it okey what i did? What do I have to do to make her look more like de original image?
Will say it can be hit or miss but when you enable Faceswap for the reference face you want to turn up the weight to at least .9 in my experience. Also might want to increase stop at a little. Maybe up to .95 but you really have to play with that slider I find. And not every attempt will be good just like with any generated images. Good quality reference image/images are important.
Does this work with fantasy characters like minotaurs, elves ,etc?
Can't say I have fully tested those but I would imagine for many it would. Faceswap may not work well if the face is not human enough looking for it to detect a face.
which base model used?
Standard JuggernautV8
Are you using an specific model or lora?
It was the default one in Fooocus which was version 8 I think at the time. No special loras.
@@KLEEBZTECH Thank you
I made a character in the past and made pictures with it, but I didn't know that I should do this, and now I want to have other angles of his face, and no matter how hard I try, it doesn't look the same. Do you have a solution?
Best thing I can think of is to try to do it a little at a time. Make one that is slightly different angle and then do another a little more. Best I can think of.
@@KLEEBZTECH thank you so much,I will try
It works well if you're creating a new face, but what it struggles with is if I have a given face and I want to create those different angles. Face swap results look awful, same on automatic1111 ReActor and FaceSwapLab
Yeah that is why I decided when creating a new person to get several angles. Definitely works much better when you have them.
Try Rope by Hillobar ... you will not find a better one. It uses multyple images from the same face to create better likeness
facing the same issue
finally: Photoshop also needed.. 😂
Well you do need some sort of image editor for things. I prefer Gimp myself.
Why do you prefer Gimp and can Gimp do skin texture? @@KLEEBZTECH .. Also, I wish they would master img2img and faceswap or look into less explored processes like headswap. Headswap is not rly mainstream yet but some things are out there on Git.
spent an hour trying to generate just the face grids with no luck, gave up
Did you change the style to line art and disable the other styles?
@@KLEEBZTECH yes I did. Went back later and tried again using different wording. Managed to get 2 that were alright then went to photoshop and flipped them around. AI truly has a mind of its own.
👍 You will get no argument from me about that. 😂
Faceswap in Fooocus doesent work
What do you mean exactly? It doesn't seem to work for you or you don't feel it works well enough?
@@KLEEBZTECH It has never worked, and even on GitHub the developers themselves admit it.
Works decent for me. I am not sure which developer you are saying says it doesn't work but I would be curious. Does it work perfectly? No. But then again really none do.