@@Animesh_12 hahah, so basiclly I would say up to 2 days, but if u giving the ai already so much info of the face it can go down, but it does take longer with 1080 ti. Also already tested it
Guys, as a synthetic media artist myself, I wished you had explained why you masked parts of the face separately, how you trained the masked parts of the face, and how you composited the face parts back together. Unless that is some sort of trade secret that you can only hint at, you ought to do a follow-up video focusing on the compositing steps.
I don't get it either, It would be cool if at least they explain why they do it that way in case they can't show the composition process for any extrange reason.
Probably training specific area's more than the general face area. So for instance you could train the eyes way longer and better than the in-build "eye's/mouth priority" option in deefacelab. Cause we tend to look at the eyes and mouth for validiation of authenticity, that's also why the deepfake probably look so good.
Great tutorial! Best I've seen. Could you explain the reasoning for exporting the different masks for each part of the face? That part was not entirely clear to me, and when do you compose them back together or use them?
I don't understand why you would use that other app to make xseg masks. DFL has a generic xseg mask you can apply and it works very well. This new method seems very tedious and takes too much time.
Definitely clean. You guys do a fair amount of compositing which helps.. That's something I haven't gotten around to learning is the post merge editing in software like After Effects.
When deepfake, 3D and AI synthetic voice finally combine in one app - oh boy... I really believe we will have our own replicas soon doing all our online communications.
Ooooo this was so sick man . 😲 Wow lorenzo's face looks a bit better than Yannick's face on jordy's body . Well done and it's also a in depth tutorial,so cool.⚡⚡
In the machine video editor, there is no GPU option showing for me, i tried with the CPU to check whether it works, but it doesn't. It's showing unable to load GPUs.
Hi! Your tutorial was very informative. I have a question. Do I have to learn the face again when I do deepfake once and then synthesize it into another video?
Editing Eye Masks. What happened to the rest of the masks? Eyebrows, face, mouth? Not sure if you just edit the training batch file and run it for each of these scenarios? before moving to the next step or if that overwrites things?
I shouldn't rap 'cos I don't know that But watching the Copycat Friday You guys know how to make my day! This deepfake was so sick! with Jordy morphing into Lorenzo and Yannick It was amazing how the faces match And just to rhyme with the last Yannick.................................... CATCH! Seriously... I don't know how to rap. I shouldn't do that! 😅
Great Informative Video Guys. We Know Final Result Took Lot Of Hard Work And Research! Kudos To You Guys And Everyone Who Was Behind The Scenes😌👏🏻 It's Officially One Of The Best Video On This Channel! 🤩✨
Great video, for the part around 10:54. It mentions about going thru the frames so the segmentation applies. When you jump forward 5 frames are you then clicking on each frame one at a time? Wasn't sure or if jumping forward 5 frames automatically loaded/applied the eye segmentation.
So I'm in step 6, and I want to train the model with Quick96 but when everything is ready suddenly stops, and says "press any key to continue" but as much as I press everything, it does nothing, any ideas?
Question: my current version has a "train XSeg" button. Can I use that instead of editing 5.XSeg) train.bat? Is there a reason why the bat edit would be better? Thank you
Hey! Very nice video! I am having a problem.. At 09:04 in the video you need to select your GPU, but mine says: 'Unable to load GPU's" Is there a solution for this?
The end result is amazing! Bravo, an incredible tutorial :) When's the mixtape dropping though?
I shouldn’t rap, cause I can’t do that 😅 thanks so much Ben!
ua-cam.com/channels/IO2LIRGYnwHkWaIP_rlXZA.html
Oh my days ben is here? 2 of my favourite creators in 1 place...what a day!
Holy crap that was clean, one of the best deepfakes I’ve seen!
true to this day
Deepfake is so fascinating, crazy to think about how good it has become in so short time. Good job guys!
Short time ?💀 He said it can take weeks if your computer is not powerful enough ....
@@Animesh_12 not rly, with a RTX 2080 I already got pretty good results after 19h
@@corpsevibes7176 🥲 gimme your system am a very big gold digger.. just kidding lol
Will it give the same result in 1080 ti also ?
@@Animesh_12 i was talking about the technology
@@Animesh_12 hahah, so basiclly I would say up to 2 days, but if u giving the ai already so much info of the face it can go down, but it does take longer with 1080 ti. Also already tested it
Wow, that's a ton of steps, but the end result is superb. Bravo!
ua-cam.com/channels/IO2LIRGYnwHkWaIP_rlXZA.html
This might be the best video you have ever made! Very informative and entertaining at the same time! Good job!
Guys, as a synthetic media artist myself, I wished you had explained why you masked parts of the face separately, how you trained the masked parts of the face, and how you composited the face parts back together. Unless that is some sort of trade secret that you can only hint at, you ought to do a follow-up video focusing on the compositing steps.
I don't get it either, It would be cool if at least they explain why they do it that way in case they can't show the composition process for any extrange reason.
Yeah. I ran mine and it worked but I had a huge box around my head
Probably training specific area's more than the general face area. So for instance you could train the eyes way longer and better than the in-build "eye's/mouth priority" option in deefacelab. Cause we tend to look at the eyes and mouth for validiation of authenticity, that's also why the deepfake probably look so good.
@@yoriettema7059 i can understand why, but it would be cool to see how they put all the parts back together..
@@yoriettema7059, yeah, it was segmented face training - something beyond the needs of most DFL users.
really clean deepfake. Beautifully done
It was cleaner than the music video. I am really impressed. The corridor crew should react to this.
ua-cam.com/channels/IO2LIRGYnwHkWaIP_rlXZA.html
Tutorials are unmatched 🔥
I was waiting for this since the moment the Kendrick released it
That deepfake was amazing!!
Also, loved that background.
I love the combination of Lorenzo and Yannick in explanation, very amazing
ua-cam.com/channels/IO2LIRGYnwHkWaIP_rlXZA.html
This is crazy insanely insane! There is no way a beginner learner can do that!
2:23 perfect sentence for training
Wow the result is so clean! It took me a while to see that you actually switched faces lol. Amazing guys!
The final result was incredible! Great work.
Great tutorial! Best I've seen. Could you explain the reasoning for exporting the different masks for each part of the face? That part was not entirely clear to me, and when do you compose them back together or use them?
It is probably to emphasize training the eyes, mouth or other parts more than the general face area.
Probably for color correction to more accurately match skin tones and to add motion blur only to face for clean transition?
I don't understand why you would use that other app to make xseg masks. DFL has a generic xseg mask you can apply and it works very well. This new method seems very tedious and takes too much time.
I was waiting for this!!
Glad you guys mentioned Corridor.
Definitely clean. You guys do a fair amount of compositing which helps.. That's something I haven't gotten around to learning is the post merge editing in software like After Effects.
who is gonna try this omg ! tons of steps
man i was WAITING for this video i knew yall wa on it the day this came out
When deepfake, 3D and AI synthetic voice finally combine in one app - oh boy... I really believe we will have our own replicas soon doing all our online communications.
4:45 the best part of the video
Lorenzo on Jordy's body looks really cool
Crazy end result though!!!!🔥🔥🔥🔥
Thank GOD there a new face extractor. I remember doing this 2 years ago and the originally face extractor was a nightmare and took FOREVER.
Wow, that's awesome. but a hell lot of effort!
This is the best deep fake I've seen on UA-cam. Very well done
The result is crazy
It looks great but the workflow is still so fricking complicated!
Never seen a copycat Friday this complex😩😩😩
I've been waiting for this! Thank you!
Ooooo this was so sick man . 😲 Wow lorenzo's face looks a bit better than Yannick's face on jordy's body .
Well done and it's also a in depth tutorial,so cool.⚡⚡
what a process!
holy jesus christ the finaly result blowed my mind away
and what do you think about the Reface app????
That software GUI is really nice and fleshed out!
In the machine video editor, there is no GPU option showing for me, i tried with the CPU to check whether it works, but it doesn't. It's showing unable to load GPUs.
Hi! Your tutorial was very informative. I have a question. Do I have to learn the face again when I do deepfake once and then synthesize it into another video?
Really a great end result. Nice work.
Love this mehn!!!🙌🏽🙌🏽🙌🏽
The Red Background Just Reminisce Me About the old Cinecom Theme ! Btw Great Rap
The result is incredible
Was waiting for this one👌👌🔥🔥
ua-cam.com/channels/IO2LIRGYnwHkWaIP_rlXZA.html
that was so smooth
The result is impressive ! Well done guys
this is the reason why i subbed
Cinecom...Finally here😁😁
Editing Eye Masks. What happened to the rest of the masks? Eyebrows, face, mouth? Not sure if you just edit the training batch file and run it for each of these scenarios? before moving to the next step or if that overwrites things?
Corridor digital that is really good deep fakes I would like to see a collab between you guys
was just looking for this
Could you please make a tutorial on audio deepfakes? I don't really know how they work.
I shouldn't rap
'cos I don't know that
But watching the Copycat Friday
You guys know how to make my day!
This deepfake was so sick!
with Jordy morphing into Lorenzo and Yannick
It was amazing how the faces match
And just to rhyme with the last
Yannick.................................... CATCH!
Seriously... I don't know how to rap. I shouldn't do that! 😅
Elitee😂
When I start the program, I get an error message, you said fill in the blanks, how do we do it?
No pop for dependencies and can't install them says not recognized as an internal or external command. Also no gpu shows up just a cpu check box
Holy cow, how long did it take you to figure out all the steps needed and to do them?? that´s craaazy.
Great Informative Video Guys. We Know Final Result Took Lot Of Hard Work And Research!
Kudos To You Guys And Everyone Who Was Behind The Scenes😌👏🏻
It's Officially One Of The Best Video On This Channel! 🤩✨
I've followed this guide but my deepfakes are really blurry even after 150k interations with batch of 8 and resolution 640. What am I doing wrong?
Have u been able to figure it out now?
the people they used look so much alike i could barley tell if it switched or not looool
😂😂😂
@14:20 did you really use a resolution of 128?
seems like you would have needed 320 or even 448 for this shot size.
you guys are tough
I was just looking for a good and simple tutorial for deepfakes!
👏🏻👏🏻 Good job guys!
Hey guys, the exports from deepfakelabs are quite blurry and lacks sharpness. I am running the filters upto 80,000+. What am I doing wrong?
Were u able to figure it out?
Great video, for the part around 10:54. It mentions about going thru the frames so the segmentation applies. When you jump forward 5 frames are you then clicking on each frame one at a time? Wasn't sure or if jumping forward 5 frames automatically loaded/applied the eye segmentation.
What of the Mac user ?
So I'm in step 6, and I want to train the model with Quick96 but when everything is ready suddenly stops, and says "press any key to continue" but as much as I press everything, it does nothing, any ideas?
All of this is awesome and proves why I should let the pros handle this 😳
ua-cam.com/channels/IO2LIRGYnwHkWaIP_rlXZA.html
Question: my current version has a "train XSeg" button. Can I use that instead of editing 5.XSeg) train.bat? Is there a reason why the bat edit would be better? Thank you
Love
should we use MVE to data_src too? or just data_dst only?
Yaaaaa I serching for this for a long time 😁😁
You are really good at this
that was so clean. Amazing
I don't see where the fake swap is??
dam that’s absolutely perfect! :0
Does this software have deep face lab integrated?
Is faceswap tool good aswell?
Woooooooow this is Amazing man
Impressive. So mush work but the result is so good. Bravo.
Do a collab with Corridor Crew feat. Film Riot and my youtube sensei are reunited now
awesome work, now I just need a 3080 to start
Wow amazing work Jordy, even though it looks so complicated to create. If I can do it, I want to deepfake my face with my cat. :)
Hold up. Did you do the deep fake? I can't tell
Big fan of you guys
THANKS! this really worked, wow dude! I like you!
а как можно с Вами связаться ?
Hey! Very nice video! I am having a problem.. At 09:04 in the video you need to select your GPU, but mine says: 'Unable to load GPU's" Is there a solution for this?
same problem, any solution? :(
Did you find a solution?
Super easy, just set a $50,000 RED camera, expensive lighting everywhere and you can do this too.
You can easily do this without all the gear we used, a DSLR and DIY lights will produce the same results.
Can you make a cg character interacting with live action like in sonic movie or detective pikachu ?
Cool thank you for the tutorial 👌🔥
The results are excellent, should call it cleanfake :). How much time did it take your computer to do the learning process ?
wow wow wow quality here
This is sick!
That looked amazing 😃😃
Can this be used on Mac?
Great job!