works great. just one Q. on the Mac M4 Pro e.g. - faceswapper works fine on all CPU cores and is quite fast.. However when faceenhancer hits in - CPUs are trotteling down and there is like 5% GPU utilization. I use only COREML and do not have CPU ticked.

Amazing Indeed! (AI) 🎉

Is this your real voice or AI voice?

ya patchwork is not recognizing me

Is it possible to install and run from an external SSD?

Do you have a video that explains what each of the options do , like "deep swapper" vs " face enhancer" etc.

bonjour j' ai ce message d' erreur in' a pas trouvé les outils(christophefanjullechuga@iMac-de-Christophe ~ % xcode-select--install zsh: command not found: xcode-select--install) pouvez-vous m' aider s' il vous plait?

Love the piano. Sure it is AI, but something about strung instruments always get to me. Great video today, Shady 🐈

Understand that this is a Research Preview. Many features are missing and there are going to be bugs while using it. I'm looking at Patchwork for its potential and not its current limitations, of which there are many. I believe that by the time this is an actual release, Patchwork will be extremely useful. More so that I already believe it to be.

I still remember my first experience with a PC in early 2000. Had no use for one prior and only got it, in order to locate / purchase music from eBay and all over the world. Well, I did that and then got into PC gaming. I was BLOWN away by what I saw and after seeing how far it has come in 25 years.... 🙀 . Midjourney Is Unkillable: Why Style Reference (SREF) makes MJ the best 👉 kzbin.info/www/bejne/a5fGnGqblq-Nbc0

Here is another tip. Run a lipsync / face enhancer swap. Then take that revised video and run it again as a target video. This time disable lipsync and do an expression restorer + face swap + enhance. Re-use your source photos from the first swap, but remove the audio clip because this time you are not doing a lipsync. You'll be amazed at how this cleans up the teeth and mouth area. Works really well with AI or 3D faces. I have not yet tested it with real people.

Try isolating some frames that show a variety of mouth and head positions of the target video. Then do frame by frame swaps, load them in a photo editor and fix the teeth for various mouth positions by overlaying the original and masking out the bad teeth from the swap. Then do another face swap with the target video but this time use the individual frames that you swapped earlier as source. Another trick is to use the face editor and open or close the mouth and try a few other sliders. That will activate the AI to do a more detailed rendering of the teeth. But it can look weird if the sliders are not right. You'll get detailed teeth but it will look unnatural. With luck you can get it close.

Could you make a video on facefusion batch setting since it recently released.

Great video. Would have liked to see how to resize windows, e.g., source and target don't need to be very large, but a larger preview window would help.

Helpful guide. Many thanks

It’s barely changed tbh what am I doing wrong

I cant seem to get anything to work on pinokio I get the same kind of error on everything something about torch and Cuda not connecting

Has anyone faced a situation where the original video is a 16-second video and the face swap only generates a 2-second video?

but how to install any github ai app? i tried gepeto ..no luck

All seemed to go ok, but upon running, I see no execution provider other than my CPU. It swaps images fine but videos aren't working (presumably because it takes forever).

FF is still seriously flawed, but when it works, it works great. Version 3 addressed a few glaring issues, batches being a major one. Thanks for clarifying everything, a few items are a bit ambiguous at first glance. The major problem with FF is that if the face shapes are not similar, the source is clipped to the outlines of the target, which can completely destroy the effect. Jawline, cheeks, the frame of the face is important. (No, tweaking the models, masks, etc. does not fix this) It's a little puzzling, as codeformer, for one example, will not do this in a stable diffusion suite like a1111 or comfy. So I usually use SD modules, but when I have occluded faces, FF shines. For whatever reason FF is excellent with occlusion. And the regional masking is great as well.

Thx a lot for the video, everything worked fine!

hi, when I try to do that it says "conda is not activated"

I just want to say to the person who recommended PINOKIO, fuck, tried it for a week and nothing downloaded successfully, except a bunch of crap!

This Job Manager should open up as a pop up widow instead of "hiding" below in the UI. It is should crucial and complex so it would be better served to be a separate window with better guidance and visuals.

Tropa do bifi 🇧🇷👉

Man kuddos to you for spending so much time. I know the struggle and your video didin't even cross 1k.

wow! do u have a discord community?

Thx 🙏

Only downloads as a still img file. How do you download video?

I live in Lenoir, about a half hours drive south of Blowing Rock. I drive 321 North quite often through Boone to visit family, go to the parkway or drive through to hangout in Valle Crucis. The night Helene hit I had considered going up the mountain just hours earlier and I'm glad I didn't. There were 3 spots between Blowing Rock and Boone on 321 that had mudslides across the road and that wasn't even the most hard hit areas. There are towns where all the shops and houses on one side of the road or another are just gone... We had several 70-to-100 year old white oak trees down on our property and were without power for several hours. We were lucky, as we don't live in a flood plain. You couldn't pay me to live in a flood plain. So many people in the harder hit areas lost everything and still need help, and lots of others lost their lives. People are still missing, bodies are still being pulled from the river. The Biden administrations response to this disaster was despicable... Sending millions... even billions to other countries around the world but won't help OUR PEOPLE FIRST. FEMA helped, to a point, but they have a reputation for being crooked by pocketing donations for "future disasters" and making a dollar in the process. It's sickening. The real heroes are the first responders, the national guardsmen, and our locals who have helped immensely in the recovery efforts.

What interest do we have in this advertisement?

zsh: command not found: conda

hi, i need to make translations of English videos to Persian and Arabic so i'm using facefusion or better say wav2lip of it, i guess the best process was to make the audio first and check if the audio sits on the file in an editing software, then i tried multiple options and i still can't make it natural, which of the settings do you suggest? Option1: gfpgan_1.4 - wav2lip_gan_96 - slow - largeSmall - BoxRegion - 0.2 - scrfd - 2dfan4 Option2: restoreformer_plus_plus - wav2lip_gan_96 - slow - largeSmall - BoxRegion - 0.2 - scrfd - 2dfan4 .

Thanks man this was very well explained. Also the pinned updates were useful to me

Can you help me swap some faces into a last supper image

Great tutorial. Tip of the hat to Fight Club :)

Why dont I have the "Windows Protected" pop up and it's downloading a "Visual Studio Installer"?

just a silly question, why your facefusion look "dark-theme"? mine is light theme and i don't know how to change it

thanks :). but I have a question, when I perform a task, temporary files are saved in AppData\Local\Temp, is there any way to change this path?

it's skipping and not overwriting all target frames when the image is small. So if the target has, let say a mustache, the resulting video will show mustache popping off and on. it also doesn't replace head looking up or down. the face flickers to the previous face. Need an option to force replace all frames.

By far the most poignant video I've seen. I traveled thru Chimney Rock twice a day and lived in Bat Cave in the cabin next door to Caulder Realty's office. Makes me so sad.

i like that tom mcdonalds is there presenting the lip sync 😀

Okay but one small problem - How do you select ONLY one of the identified twins? I did tried everything but it still detected and changed both faces instead the one I would like to change while other one kept unchanged.

Thanks man.

Am I the only one that wants to take the Udio a.I music and put it my fl.studio and create a song my own lyrics

I've managed a few videos with Face Fusion but the majority of them just hang with "ERROR" in the output window either during the initial processing phase or later in the process. I can't see what is different between the ones that work and the ones that don't.

Your Videos are top, it helps me a lot, thank you for that. I would be interested to know what is actually important in terms of hardware properties for AI applications such as: generating 50min 4K videos in FaceFusion. I am very excited about what is possible with Pinokio and I will get an Nvidia graphics card. So what is the most important thing, as many CUDA cores as possible? Pixels or texture rate? VRAM as high as possible, 48 or 80 GB best? Memory bandwidth? I don't want to do this as a hobby but as professionally as possible.

thanks

Access to model black-forest-labs/FLUX.1-Redux-dev is restricted. You must have access to it and be authenticated to access it. Please log in?