Major Ass Kicking Video. Your name throws me off a trifle, but then again there are folks who name themselves "Good guy this-or-that" and well is nothing but the opposite. Really enjoyed your tutorial and the way you present things as well as your sense of humor. I kept thinking "Fight Club" and was expecting a flash of "skin" LOL. I look forward to more. As far as TH-cam goes, you are the authority on FaceFusion. Thank you for the enlightenment! Subbed.
@@noNumber2Sherlock thanks! I'm glad you liked it and appreciate my brand of humor. Fight Club was definitely the inspiration for this one. I just didn't put the crying children in it.
Have been sitting here and trying to think of something witty about how I have been programmed Manchurian Candidate style, by the Steve Jobs subliminal images. However, I am ever distracted by how powerful a program this is. Great tutorial, as per usual, so don't come at a Kitty with how you stumped my ability to be silly 😾 PS> I actually already knew what a .json file is 😸
Now I can't get Sinatra music out of my head for some reason... Yeah, I'm pretty excited about the new feature and this is just a small part of what I used for the video of what I told you I'd been working on for quite some time. I guess I need to go watch Army of Darkness now. It's been quite a while.
Thanks for you great tutorial. I find the "niche" possibility of editing the json files very interesting and tried to create a file allowing me to create several videos with the same source picture. I first tried with a two step job then copied the corresponding json file from the completed folder to the queued one, edited it and changed the name of the two target videos (in my case just a change of number, everythingelse the same). I also changed the status from "completed" to "queued" because if I let it like this it would fail. However, despite of doing this, I still get a "fail" after having restarted FD and ran the job. Anything else I should edit in the json file in order to have it run? Thanks in advance!
I realized that it probably cannot be done since first the target is getting analyzed to determine the output video resolution and fps. If you copy and paste this information, in probably cannot work since every target video is different. Thanks anyway!
@@jplon8926 yep, it sounded like a good idea, not really possible. At least not to where it would be worth the time consumption. It would be pretty easy to do the opposite though. Where you make copies of the same step over and over for multiple sources (faces) to swap onto the same source.
It's not necessary, but it doesn't hurt at all. However, if your images that you have of a certain individual aren't the greatest because they don't show much definition or they are bad angles or bad lighting, then it helps to have other options to "average" the features together. I've just gotten in the habit of having multiple images of many of the peoples' faces that I use over and over. But no, having multiple angles will do nothing to help with a target that has bad angles.
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.
Yes, definitely works with images. I just never show images in my videos as, well, first of all, it's never as exciting to see the output and I very rarely have a use for single image outputs.
I loved this queueing option to be honest. Very helpful. By the way, may be I missed it but, the functions in the "facefusion next" are different than 2.6.1. 1-) So is it only in developer version? 2-) is there a way to have developer version or it is something unique to few people only? 3-) Is it possible to have also a matte for the swapped face to have more control over the process? Thanks in advance.
1. Yes, I did state towards the beginning that this is only currently in the NEXT developer version and will be available to everyone when 3.0 is released. 2. Anybody can test out the NEXT version with the understanding that it is a developer release and that things constantly change and can break and you are using it to test and report any bugs you come across. If you are interested in trying it out, check out the FaceFusion Discord and ask in the "early-adopters" channel. 3. Nope, not possible. Thanks for watching!
Nice video, as usual. I wonder 2 things:1) why on Discord we are more than 30k and they are not all subscribed to your channel? It should be mandatory. 2)I am waiting to see a video with which you use my graphics🤪 if you would like to do so, let me know I will optimize the Patch and you can show how to apply it. Let me know.😅😅😅
Oh, believe me, I would love this. Getting barely 200 subs in a month when 30,000+ are using this software is definitely frustrating. Oh well, hopefully that will change soon. Especially with the video I have been working on for some time now. But I can't release it yet. Blame Henry :P I assume by your graphics, that you mean the interface changes that you created for FF, I would love to, but I, myself, even use a very different looking interface most of the time, but I try to use the default so that viewers aren't confused by any changes and can easily relate what's in the video to their own. As always, thanks for the support!
@@shadyendeavor I meant just that, you could show users com and install mods. (e.g. my graphics mods) or the various other interesting mods found on the discord channel.
Hello, thank you for you high quality content ! Im a nooby in the AI word and with tools like Facefusion. I managed somehow to make it work on mac M1 but the problem is that when i can't download videos that are more than 1 or 2 seconds. And even for the 1 and 2-second video i have to wait almost 20 minutes. Do you know what i could do to fix that ? Again i'm a nooby i kind of need explanation like for a kid if I may say.
First off, welcome to the AI world. It's super crazy, but can be a lot of fun. Second, I'm not entirely sure what you mean by not being able to "download videos". I'm not sure if you mean the actual downloading of the video once it's finished to your hard drive or if you mean the actual creation/running of the face swap or what. But here's what I'd recommend. I have a lot of videos on FaceFusion and I'll link the playlist I have of all of them at the end of this comment. You should start off with the "Basics" videos (probably in order) and then if you want more information, check out the "definitive deep dive" video. That should be more than enough information to really get your on your way. And yes, I'm on an M1 also and it can be very very slow. Just realize that any of the processing in my videos, I'm usually speeding it up (or cutting it completely) in the edit to not bore you and the other viewers. If you still have questions, feel free to ask. Thanks for watching! th-cam.com/play/PLMwq9QXPN0iDReOut5vKekXT4V6wGTET8.html
@@shadyendeavor Yeah, what I meant is that I can't create the video if the target video that I pick is longer than 2 seconds. So for a video of 60 second i literarily have to do the hole thing 30 time… It's been 2 entire day, I'm fighting to find a solution, but I'm terrible at it. I even bought the paid app Facefusion, I thought maybe it would work better, but it definitely doesn't. I have no problem waiting 1 or 2 hours for 1 video of 1 minutes, but at least it needs to take the full 60 seconds at once. 2 seconds per 2 seconds is a nightmare. I can't figure out what is wrong with my situation and why it does that. But to be honest there is a chance that I'm missing something, but what…
@@paullaurent4128 It really all depends on your settings. Make sure you have "coreml" selected for your Execution Provider. Make sure you aren't using "Frame Enhancer". First try things with only Face Swapper turned on for the processors. Don't increase the pixel boost or anything. Keep all of the defaults. Then it depends on the resolution of your file. Even on my M1, using 4K video, I'm getting a couple frames/second with higher settings than the defaults. Be sure you have your terminal open so you can see the progress and that's where you can see how long and slow/fast it is processing. Don't look at the "output" window. The time there means nothing. Either way, everything I've said here is stuff that's in my videos. ;)
@@shadyendeavor I did all you said but it still doesnt work. I dont get it. It cant work with video longer than 1 second. I have tried to compress the files to get light files. The weirdest thing is that I have used heavy fils of one second with face_swapper and face_enhancer and frame_enhancer togethher that is 23 Mo and it works in only 3 minutes. but when I use a file of 5,8 Mo that is 10 seconds so the file is lighter it still doesn't work. And i have used only default setting with coreml and no boost and no proccessors. So it doesnt looks like the problem is how heavy is the file or how much proccessor i use. Only 1 seconde video work, i dont get it. Maybe I could show you my screen if you have time of course. Sry to bother you with my problems.
If I have a current job running with a bunch of steps in queue pending, can i still start a new job and run it? if so, will the new job start automatically once the current job is finished?
Nope, while a job is running, you pretty much can't do anything with your current FaceFusion environment. Now, I said it that way specifically, as you could have another one running (either another conda environment or Pinokio) and work on making another job with that one. Not ideal, kind of a PITA, but yes, it's possible...sort of. I just wouldn't recommend it.
Yes and no. I've had times where I've had an image or two that didn't really make the swap look like the person, so I added another 1 or 2 and then it looked much better. I played with those a bit and yeah, some single images just aren't as good at portraying the subject. So, as I do this so much, I just got into the habit of having multiple images for ever subject I use. At least if it's not someone new that I want to try really quickly. I'm sure a lot of the time, it's not making much of (or any) difference), but I don't think it hurts in anyway. So why not? ;)
@@AbhijeetMaurya-g1l colab is not supported by FaceFusion currently, and hasn't been for a while. If you want to use it for free, you'll need to install it locally on your computer.
cd facefusion (make sure you're in your FF directory) conda activate facefusion (make sure you're in your venv) python run.py or python facefusion.py run (if you have NEXT version)
Yep, Gradio is not a great UI. And it's definitely on the Roadmap to be replaced, but it's a huge venture to switch over. But yes, could definitely be better.
For some reason, the Remix is not working for me. I just end up with 2 files, 2 separate face swaps, as if I'd done add-step, but I'm definitely saying Remix. Any ideas would be greatly appreciated from anyone! Thanks for your videos!
@@BobDoyleMedia I'm really not sure what could be the problem. But just to make sure, you definitely did add-step for the first swap and then remix-step for the second swap? And with nothing else done in between those steps?
![I Replaced ALL my ADOBE APPS with these [free or cheaper] Alternatives!](http://i.ytimg.com/vi/5EfqHg49kMk/mqdefault.jpg)
Major Ass Kicking Video. Your name throws me off a trifle, but then again there are folks who name themselves "Good guy this-or-that" and well is nothing but the opposite.
Really enjoyed your tutorial and the way you present things as well as your sense of humor. I kept thinking "Fight Club" and was expecting a flash of "skin" LOL.
I look forward to more. As far as TH-cam goes, you are the authority on FaceFusion. Thank you for the enlightenment! Subbed.
@@noNumber2Sherlock thanks! I'm glad you liked it and appreciate my brand of humor. Fight Club was definitely the inspiration for this one. I just didn't put the crying children in it.
@@shadyendeavorLOL Yes now I'm 100% certain you are badass. ( was 99.9% prior ) **Gentleman's Nod** towards your way.
Great tutorial. Tip of the hat to Fight Club :)
Have been sitting here and trying to think of something witty about how I have been programmed Manchurian Candidate style, by the Steve Jobs subliminal images. However, I am ever distracted by how powerful a program this is. Great tutorial, as per usual, so don't come at a Kitty with how you stumped my ability to be silly 😾
PS> I actually already knew what a .json file is 😸
Now I can't get Sinatra music out of my head for some reason...
Yeah, I'm pretty excited about the new feature and this is just a small part of what I used for the video of what I told you I'd been working on for quite some time.
I guess I need to go watch Army of Darkness now. It's been quite a while.
Thanks man.
Thanks for you great tutorial. I find the "niche" possibility of editing the json files very interesting and tried to create a file allowing me to create several videos with the same source picture. I first tried with a two step job then copied the corresponding json file from the completed folder to the queued one, edited it and changed the name of the two target videos (in my case just a change of number, everythingelse the same). I also changed the status from "completed" to "queued" because if I let it like this it would fail. However, despite of doing this, I still get a "fail" after having restarted FD and ran the job. Anything else I should edit in the json file in order to have it run? Thanks in advance!
I realized that it probably cannot be done since first the target is getting analyzed to determine the output video resolution and fps. If you copy and paste this information, in probably cannot work since every target video is different. Thanks anyway!
@@jplon8926 yep, it sounded like a good idea, not really possible. At least not to where it would be worth the time consumption. It would be pretty easy to do the opposite though. Where you make copies of the same step over and over for multiple sources (faces) to swap onto the same source.
Indeed sleep is good! Facefusion has ruined my sleep hahah
Sleeping's my favorite! And Jobs definitely helps with it.
1:07 how did you swap side view without flickering?
That's the background video that doesn't have any swap in it. Same with the man sleeping right after. No swap.
Hi Shady, I saw you droped a couple of pictures for source, are these a person's face with different angles? Is it helpful for face swapping?
It's not necessary, but it doesn't hurt at all. However, if your images that you have of a certain individual aren't the greatest because they don't show much definition or they are bad angles or bad lighting, then it helps to have other options to "average" the features together. I've just gotten in the habit of having multiple images of many of the peoples' faces that I use over and over. But no, having multiple angles will do nothing to help with a target that has bad angles.
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.
Preview would be good like After Effects Ram previews
thanks
Very helpful ! Is this working on Images also or just videos?
Yes, definitely works with images. I just never show images in my videos as, well, first of all, it's never as exciting to see the output and I very rarely have a use for single image outputs.
ancient invention 😂 anyway, you explained it so well!
Thank you! I'm glad it made sense. It always makes sense in my own head, but I never know how people are going to react to it from their perspective.
I loved this queueing option to be honest. Very helpful. By the way, may be I missed it but, the functions in the "facefusion next" are different than 2.6.1. 1-) So is it only in developer version? 2-) is there a way to have developer version or it is something unique to few people only? 3-) Is it possible to have also a matte for the swapped face to have more control over the process? Thanks in advance.
1. Yes, I did state towards the beginning that this is only currently in the NEXT developer version and will be available to everyone when 3.0 is released.
2. Anybody can test out the NEXT version with the understanding that it is a developer release and that things constantly change and can break and you are using it to test and report any bugs you come across. If you are interested in trying it out, check out the FaceFusion Discord and ask in the "early-adopters" channel.
3. Nope, not possible.
Thanks for watching!
@@shadyendeavor thank you so much! Appreciate your time and effort. 👍
Nice video, as usual. I wonder 2 things:1) why on Discord we are more than 30k and they are not all subscribed to your channel? It should be mandatory. 2)I am waiting to see a video with which you use my graphics🤪 if you would like to do so, let me know I will optimize the Patch and you can show how to apply it. Let me know.😅😅😅
Oh, believe me, I would love this. Getting barely 200 subs in a month when 30,000+ are using this software is definitely frustrating. Oh well, hopefully that will change soon. Especially with the video I have been working on for some time now. But I can't release it yet. Blame Henry :P
I assume by your graphics, that you mean the interface changes that you created for FF, I would love to, but I, myself, even use a very different looking interface most of the time, but I try to use the default so that viewers aren't confused by any changes and can easily relate what's in the video to their own.
As always, thanks for the support!
@@shadyendeavor I meant just that, you could show users com and install mods. (e.g. my graphics mods) or the various other interesting mods found on the discord channel.
Hello, thank you for you high quality content ! Im a nooby in the AI word and with tools like Facefusion.
I managed somehow to make it work on mac M1 but the problem is that when i can't download videos that are more than 1 or 2 seconds.
And even for the 1 and 2-second video i have to wait almost 20 minutes.
Do you know what i could do to fix that ? Again i'm a nooby i kind of need explanation like for a kid if I may say.
First off, welcome to the AI world. It's super crazy, but can be a lot of fun.
Second, I'm not entirely sure what you mean by not being able to "download videos". I'm not sure if you mean the actual downloading of the video once it's finished to your hard drive or if you mean the actual creation/running of the face swap or what.
But here's what I'd recommend. I have a lot of videos on FaceFusion and I'll link the playlist I have of all of them at the end of this comment. You should start off with the "Basics" videos (probably in order) and then if you want more information, check out the "definitive deep dive" video. That should be more than enough information to really get your on your way.
And yes, I'm on an M1 also and it can be very very slow. Just realize that any of the processing in my videos, I'm usually speeding it up (or cutting it completely) in the edit to not bore you and the other viewers.
If you still have questions, feel free to ask. Thanks for watching!
th-cam.com/play/PLMwq9QXPN0iDReOut5vKekXT4V6wGTET8.html
@@shadyendeavor Yeah, what I meant is that I can't create the video if the target video that I pick is longer than 2 seconds. So for a video of 60 second i literarily have to do the hole thing 30 time… It's been 2 entire day, I'm fighting to find a solution, but I'm terrible at it. I even bought the paid app Facefusion, I thought maybe it would work better, but it definitely doesn't.
I have no problem waiting 1 or 2 hours for 1 video of 1 minutes, but at least it needs to take the full 60 seconds at once. 2 seconds per 2 seconds is a nightmare.
I can't figure out what is wrong with my situation and why it does that.
But to be honest there is a chance that I'm missing something, but what…
@@paullaurent4128 It really all depends on your settings. Make sure you have "coreml" selected for your Execution Provider. Make sure you aren't using "Frame Enhancer". First try things with only Face Swapper turned on for the processors. Don't increase the pixel boost or anything. Keep all of the defaults. Then it depends on the resolution of your file.
Even on my M1, using 4K video, I'm getting a couple frames/second with higher settings than the defaults. Be sure you have your terminal open so you can see the progress and that's where you can see how long and slow/fast it is processing. Don't look at the "output" window. The time there means nothing.
Either way, everything I've said here is stuff that's in my videos. ;)
@@shadyendeavor I did all you said but it still doesnt work. I dont get it. It cant work with video longer than 1 second. I have tried to compress the files to get light files.
The weirdest thing is that I have used heavy fils of one second with face_swapper and face_enhancer and frame_enhancer togethher that is 23 Mo and it works in only 3 minutes.
but when I use a file of 5,8 Mo that is 10 seconds so the file is lighter it still doesn't work. And i have used only default setting with coreml and no boost and no proccessors.
So it doesnt looks like the problem is how heavy is the file or how much proccessor i use.
Only 1 seconde video work, i dont get it. Maybe I could show you my screen if you have time of course. Sry to bother you with my problems.
If I have a current job running with a bunch of steps in queue pending, can i still start a new job and run it? if so, will the new job start automatically once the current job is finished?
Nope, while a job is running, you pretty much can't do anything with your current FaceFusion environment. Now, I said it that way specifically, as you could have another one running (either another conda environment or Pinokio) and work on making another job with that one. Not ideal, kind of a PITA, but yes, it's possible...sort of. I just wouldn't recommend it.
@@shadyendeavor Got it. Thanks for the clarification!
do you feel it makes any difference at all whether you have one source image or you have several?
Yes and no. I've had times where I've had an image or two that didn't really make the swap look like the person, so I added another 1 or 2 and then it looked much better. I played with those a bit and yeah, some single images just aren't as good at portraying the subject. So, as I do this so much, I just got into the habit of having multiple images for ever subject I use. At least if it's not someone new that I want to try really quickly. I'm sure a lot of the time, it's not making much of (or any) difference), but I don't think it hurts in anyway. So why not? ;)
hello sir I am using the facefussion on the google clob but now its not working can you tell a how to excess it and if possible it should be free one
@@AbhijeetMaurya-g1l colab is not supported by FaceFusion currently, and hasn't been for a while. If you want to use it for free, you'll need to install it locally on your computer.
How do i start the program when it’s already installed on conda, I’ve forgotten how the commands
cd facefusion (make sure you're in your FF directory)
conda activate facefusion (make sure you're in your venv)
python run.py
or
python facefusion.py run (if you have NEXT version)
This is great. But what is stopping FF to also transfer hair color and definitely the skin color without which face transfer can not be complete.
just a silly question, why your facefusion look "dark-theme"? mine is light theme and i don't know how to change it
@@DarienLingstuyl it's automatic based on your system/browser's setting
@@shadyendeavor i have facefusion installed on pinokio, do you how can i make it dark theme there?
@ Pinokio > Settings > Theme > dark
😉
Boy does this interface need improvement - something like Adobe's media encoder. Otherwise this option is nice to see.
Yep, Gradio is not a great UI. And it's definitely on the Roadmap to be replaced, but it's a huge venture to switch over. But yes, could definitely be better.
For some reason, the Remix is not working for me. I just end up with 2 files, 2 separate face swaps, as if I'd done add-step, but I'm definitely saying Remix. Any ideas would be greatly appreciated from anyone! Thanks for your videos!
@@BobDoyleMedia I'm really not sure what could be the problem. But just to make sure, you definitely did add-step for the first swap and then remix-step for the second swap? And with nothing else done in between those steps?
👌👍👍👍
not working for me