FaceSwap comparison, which is the best method in Fooocus?

🔥I will be comparing all the face-swap technique possible in Fooocus. We'll explore all methods to find the best method.

👉Discover the power of Fooocus AI in face swapping. Watch until the end to see the amazing results!
 
#fooocus #StableDiffusion #DigitalArt #ImageManipulation #CreativeTools

00:00 Introduction
00:28 Using Image prompt in Fooocus
02:01 Using Image prompt plus Inpaint Fooocus
03:43 Using Image prompt plus variation in Fooocus
05:00 Using FaceSwap and PyraCanny in Fooocus
06:23 Comparing result, Perfect FaceSwap technique
06:57 FaceSwap with different poses
08:08 Final thoughts.


Older video using image prompt and Inpaint

Older video using image prompt and variation (client order completed)

Playlist For Fooocus AI - Stable Diffusion

Playlist For ComfyUI - Stable Diffusion

Image Generation Tricks for Automatic1111

[Dennis-bfll]

I made facegrids of different faceangles.
Used Imagepromt and put stop at up to 1.0.
Then uploaded the face grid I wanted and choosed pyracanni with weight at 0.766

Then i wrote a prompt about keeping facelines, eye colour etc.

I let foocus make 32 picture each time.

After 3 -4 tries you will have an acceptable Image.

I made this for the poses:
Frontal, half left, half right, full left, full right and also each of them with slightly looking up/down.

Now you have enough material for every face direction.

At the moment I am making each posiotion with emotions (happy, sag, thinking, shy etc).

When I am ready I will start an AI influencer

[edinsonriveraaedo]

thank you for the tips. I have sometimes problems and it wasn't clear for me how to fix it.

[kromecrowell]

How to train a Lora for free would be so great

[ximorro]

Thank you very much! This is a difficult task. I reached the conclusion that maybe you can trick the viewer because the AI can get a very similar face and if you don't know the person it's probably enough. But trying to use faceswap (alone or together with other methods you explain here brilliantly) with people you know like yourself, friends or family... you can notice the generated image is not the same person, it's like an alternative version of that person but not that person you know, a very unpleasant uncanny valley effect!
Maybe training a Lora, as you point out, is the only way to get a consistent faceswap of a real person...

[lorenakademar]

as usual..very interesting! Have a great weekend

[adapt-or-die-trying]

Yes, please show us how to train a lora of one model to get her face to show consistently across different angles 🙏

[nicdisan]

Spectacular tutorial, thank you very much for what you did for the community, but maybe we need to try with loras to define the last faces that look in different directions. I stay online for other experiments. A question do you think there will be in fooocus the implementation with FLUX? thanks

[HumzaYousafFilms]

what are you favorite models for realism? face swapping

[rikkles]

would love a video explaining how to train a Lora!

[TomSmith-yhju]

based on my graphiccard, the programm downgraded me on default settings, and with this, I couldn`t get a high detailed background

[MMAHypeWatch]

Can't you take a picture of someone and then mask their face, generate a prompt and let it build a picture around that persons face???

[roumerh]

Please what checkpoint and Lora did you use

[joechip]

You swapped faces from photo to photo. Is there a way to swap the face from a portrait photo to a certain drawing style, e.g. a cartoon style. Can this also work with in paint? Say I have a drawing of a cartoon character and want to replace it's face with a cartoonish version of a real person's photo. What I want to achieve is blending faces of children into drawn fairy tale images, so the child can read a fairy tale where the hero is he or she themselves.

[RafaelSequera]

fooocus works well with AI generated faces, but it wont work with real faces, and thats it for now...