Update: I've noticed several comments of people asking where to get access to DFM model files to use with Deepface Live. I make free models regularly, which you can access here: drive.google.com/drive/folders/1tH-8SXb5R4-W1c3wIVXy4aYnZOP0Om-1?usp=sharing ------------------------ I apologize in advance for the weird sounding voice audio. Apparently OBS was recording both my XLR mic along with the mic built into the Logitech Brio.. the result is a bit.. unsettling, lol. *Update* I found a walkthrough from the developer (Iperov) about what settings to use to create your own 224 res model that you can import into DeepFaceLive: github.com/iperov/DeepFaceLive/blob/master/doc/user_faq/user_faq.md
Thats cool. I expected some latency. Sure you dont want to step up your shock value? The potential of this technology is really interesting. Cant wait until you upgrade.
At 11:45 which version of deepface lab is that? I don't have "6) export SAEHD as DFM" I already trained a bunch of models with Quick96. Do I have to start over and train with SAEHD so I can export DFMs?
Yes, you'd have to use and SAEHD model to export as DFM, so a Quick96 will not work. That said, you should have that batch file in your folder if you're using any of the last couple versions of Deepface Lab.
@@BeardofBeesPool Tested my Arnold model. It looks "ok" but it's not as good as I thought it'd be. I'm kinda curious how much they trained the models that come with DeepfaceLive.
Hey druuzil, i have a question, is that normal that i got very low fps when i use deepfacelive + playing a game ? i went from 30/50 fps with dfl alone, then 8 fps when playing + recording a game + dfl
There seems to be some debate on that. There's an old version of DF Lab where you can supposedly use a DFM file (a model exported into the format that works in DF Live) and put that back into regular DF Lab, then apply it to a video clip like you would normally do with DF Lab. The problem is you can't train the model any further that way, and usually when you use a new video clip (like if you want to put Tom Cruise into an episode of "Friends" for example), it may not work that well because usually you'd have to train your model against that new video clip for a while before the final merger process. I hope that made sense. Anyway, I have yet to see someone do it, but someone was claiming in Discord earlier this morning that there was a process to do it, so I plan to test it out at some point. If it works, I may do a tutorial on it.. but generally, making a model of your own would offer superior results for a few technical reasons.
Thanks. I am wondering though, will this work for all websites that use Webcam? Will it always show the live deepfake face of me, not the real me face?
I don't know, I haven't tested it outside of the DF Live environment. When I livestream it via OBS, it's just OBS capturing the screen and sending it to UA-cam. I'm not sure how a chat website or something similar that uses the webcam would be able to interface with DF Live.. Something I'd have to play with.
Nice video. Just to add to it cos you kept mentioning resolution and I noticed yours was in default setting. I normally increase mine to 960 and it’s fairly clear and detailed all round
I just need to ask that can I select my own model ? like my friend or some other person who is not common. because there is no any choose file option for getting my own known ones photo saved in my laptop. I want to deface other person photo. please help me
You can create your own models if you have proper hardware and are willing to put in the time/effort. I have tutorials on this channel that go over the process. It is somewhat complicated though and as mentioned, time consuming.
How are you liking the WX6800? planning on getting myself one as im currently on a Titan V and its doing fine but i would like some more umph and id be able to get the W6800 for a nice price if i can sell off this Titan GPU.
I'm having problems with it actually. Trying to do an RMA with AMD to get a replacement. Not 100% sure if it's the card, AMD's drivers, or Deepface Lab to blame though. Trying to get another AMD card locally so I can test and see if the problem is AMD in general or not.
This could actually be very useful in corporate environments. Imagine people working from home in a zoom call, unshaved , or without make up ( women of course) can now enter a zoom call with their office face. There would be a market for it I guess. Or UA-cam presenters that want to look a bit different or better online can rework their face digitally and iron out the imperfections for their online presence
Yeah I could see that. Or live interviews etc for celebrities who want to look their best but don't want to have a makeup artist spend time/$ on making them look good. Could probably be used in a lot of ways.
If you watch his videos (I'm a big fan of his too), the models are pretty bad quality and don't really look to me like they came from DeepFaceLive.. Maybe FaceSwap? Did you watch his video "Star Boat"? Hilarious.
@@MagnaRads I tried messaging him a few times on his channel to see if he'd let me help improve the Deepfakes but didn't get a response. I think in fairness the bad quality actually makes it funnier anyway.
I can only assume you didn't use enough src material that has the person talking, or it's an issue of how you trained the model. Did you follow any of the tutorials on my channel?
As far as "using it on other apps" goes, I think that pretty much requires using something like OBS and a virtual camera to "clone" the DF Live output into the app you're trying to use it with. I've had limited success with that though and haven't screwed with it much in quite a while. I tried using it with Omegle but wasn't able to get it working. As for what kind of hardware you need to run it, you need a dedicated GPU (if your laptop only has 4GB of system memory I guarantee it's not fast enough to run this).. I think the recommended specs are something like an RTX 2080 Ti (or RTX 3070 which is equivalent) or faster. That'd be to run 224 resolution models.. 320 resolution (which of course looks better) requires even faster hardware. It's not a filter, it's literally deepfaking you in real time, so it's unfortunately quite system intensive.
If we have lower specification laptops. In that case what we need to do. Can we use cloud gpu providers like(papersspace). Any suggestions will be help full.
Because it's in the cloud..? You can't run a local application like DF Live in the cloud and interface with your local webcam without extreme latency I'd imagine.
@@RavulapalliNareendra There is delay, yes. It's like a remote desktop session where the game is being played on the server and you just see it on your screen.
Great video!!! I love how clear you make these explainations. I've got one question though. How can I create my own model for deepface live and how can I add it to deepface live? Okay that's 2 questions. I apologize
I am having problem swapping the face even after I select a picture I want to swap. It’s only source frame and aligned face . It’s not showing face swapper or final frame. The deep face live output isn’t showing as well. Pls can you help me out.
What are you talking about. There are no graphics driver links in the description. DF Live is definitely still available at least on the MEGA download link.
Hello, why i have no sound?....i record a video with obs, then i deepfacelive it but there is no sound....at the window out, not even when i record with obs.... why? thnnks!!!
if you're asking about using other models than the ones that come with DF Live, yes you can do that. I have a full tutorial on my channel on how to make your own, and I supply free models all the time. Check any of my newer livestreams for a link to my Google drive to download free models.
512 really isn't viable in Deepface Live. Unless you have an RTX 4090 video card, it basically won't run at all, and even if you do have a 4090 the frame rate is pretty bad. 320 is much more realistic. Do you know what video card you have?
@@DruuzilTechGames i made a 512 model of my wife when she was younger :) it brought so much memory back for us with tears flowing freely of course on her side.
Brother thank you so much for you video. Could you please guide me on the part where I actually wish to use this in a live video conference format. For example, in Discord. How do I export the DFL picture into a different program?
It involves using OBS and virtual camera, but I've not used it for that personally. You'd probably have to ask in the DeepX discord channel.. there's a couple guys in there that know how to do it, I just haven't had much luck with the virtual camera stuff.
Can you share how to bust someone using this software? This is obviously popular among scammers, does the face de-sync if it's obstructed? What needs to be covered? Half the face? The eyes? It's obvious if the face is turned to the side that it will get distorted, what other ways can you break this software?
It definitely doesn't handle facial obstructions well, unless the model is specifically trained to help with that which is time consuming and most people don't bother to do it. Unless the person does something specific during the video to destroy the illusion however, idk. PIxelation around the eyes, inside the mouth... facial distortion at certain angles if the model didn't get properly trained or the faceset was insufficient to cover usage at all angles.
In DF Live, use the "Merged Output" window (so you can see yourself as the deepfake character in a larger window), then in OBS do "Window Share" and select that Merged Output as the window you want to share.
@@DruuzilTechGames thnks !! the window thing i cliked but did nothing... ill try the source choose the merged window as the source for obs (if the bigger windows still doesnt appear) ill try later thanks for the quick answer !
It's pretty similar. If you go to the links I provided in the description of the video, you can find a tutorial on the creation of a DF Live model. Basically you still make the model in DF Lab, at 224 resolution. You need a good source faceset of varied images (this is the character you want to see yourself as in the live broadcast). Needs various lighting conditions and facial expressions, looking in all directions. You then train that set of faces against the "RTM" faceset (found via the "Magnet" link on the Github site, again link in the description) using the instructions for creating a Deepface Live model (again, linked). Once you've trained it sufficiently (mine started to get decent around 1.6-2m iterations), you export a "DFM" file and place that into the DF Live models folder, after which you can select and use that model in DF Live.
@@DruuzilTechGames I a core i7 720q and an nvidia gt 330m installed dircectx 12 when I check on my nvidia panel it says directx 11.0 but on dxdiag it says directx 12 so I taught it should be able to run the directx 12 version but I get the openCL library error when the nvidia driver is not installed but now that’s it installed I’m getting a different error
Not sure what you mean by uploading a picture. Upload to what? Deepfakes require thousands of images, and a fairly long period of training time to get the model to work properly. I have tutorials on my channel that cover how to do it if you're interested in learning.
DF Live runs locally on your PC. I don't think you could run it in a cloud environment, and there'd be a ton of latency if you could. Additionally, Colab banned using Deepface Lab, so I doubt they'd support DF Live either.
You can't. You can only export the individual images, which you then would have to import into something like Adobe Premiere Pro (as an "image sequence") to convert it back into an mp4. It loses the audio too, so you'll need to pair up the audio track with that in Premiere to export the final product.
Well you can do that if you want to make your own models using Deepfacelab. Better results too than DF Live. Just much more complicated and time consuming.
Download the DFM files, then go into the Deepface Live folder, then the models folder and put the DFM files there. Then you can select them while running the DF Live application.
Yes. I have a tutorial on my channel that covers how to do it. You can also download the ones in my Google drive which should be in the video description.
There is no "uploading" to DF Live. You just download any of the DFM files in the drive, and put them in the DFM folder inside the DF Live main folder. Done.
Please check out some of my newer videos on the subject. I did a full tutorial on how to create a model for Deepface Live. DF Live and DF Lab are 2 separate software suites. You create the model in DF Lab, and export what's called a "DFM" file, which you can then use in DF Live.
Hello… I’m running on 8GB rtx 2080 laptop It won’t give me good results. I’ve seen people run it on 2070 and it works well mine is shaky and looks very unreal
Has nothing to do with the GPU, and likely something to do with how you're training the model.. Like what settings, when you're enabling them, for how long, etc. *edit* Are you referring to how well the models that come with DF Live are running/look on your system vs his? Are you using the same models the other person did? If not, it can vary from model to model how well they perform based on how they were trained.
@@DruuzilTechGames yes… I’m referring to how well the DFM are looking. I didn’t train them myself.. I Download one out of yours and it was a disaster. Do you think it has to do with my pc’s camera?
It has to be some kind of settings, and maybe the lighting in your room. Soft, white light tends to work best. You can find me on Discord at "Druuzil_TG" if you want to send a pic.
Yeah but you have to create the model that way in the first place, and I'm pretty sure you can only do it with Head models, not "Whole Face" models. Even then, it doesn't work super well unless you remain fairly still/don't move around.
@@aam705 Is that a laptop? I don't think the 1650 Ti is a desktop card. Anyway, that's a pretty low spec machine to be even attempting something like this. To give you an idea of the systems I'm using and still only getting ~30 FPS at 480p, I have an i9 11900k/RTX 3090 in one system, and a Threadripper 3960x/AMD W6800 Pro in the other. Both get similar results.
Real time voice changing isn't super good yet. There's like 1 or 2 applications out there but they're not very good quality and there's quite a bit of lag because it has to change the voice on the fly.
@@DruuzilTechGames yes flickering... I just got the PC Alienware x15 r2 3070ti 8gbvram... I tested it with a dfm(Keanu reeves) downloaded on the dflive. Would I be able to train a model with the PC?
idk why people keep buying Alienware laptops to make deepfakes, lol. I've seen so many people buy similar systems. Laptops are not a good idea for training models 24/7. They perform poorly vs desktops and get very hot. Anyways, the 8GB of VRAM is going to limit you. You "may" be able to train the 224 RTT model, but probably more like 192. Not ideal.
Hah yeah that'd be a good one for me. I'm gonna see how well the Arnold model I'm training ends up working and then I may do that. So many Deepfakes.. so few computers...
Non-celebs are not in the public eye and are not viable without express permission. Besides, part of making UA-cam videos is so they'll get clicks. If nobody knows who the person is, nobody is going to watch it.
Please have a look at the tutorial videos I've made on how to make a Deepfake. Basically though the RTM faceset is about 63k images in a pak file that you can train your model against so it learns a ton of faces, lighting conditions, angles, etc.
@@jaya_surya_a5118 If you don't have source images you can't make anything. The RTM faceset cannot be used to make a regular video. Please go checkout the 2 tutorial videos I made on my channel which explain 100% how to make a Deepfake from scratch. One explains how to do a regular model, one explains how to make a live model.
I guess it depends on what you're trying to do. If you want to livestream with DF Live, use OBS (and OBS will allow you to select an audio source). If you're trying to apply a deepfake to a video clip using DF Live, then it doesn't add audio and you'll have to use some kind of video editor like Premiere Pro to add an audio track back in after the fact.
Hi, great video, I can´t find the way to train my own face, I extraxt the face but where have to train it ? I have installed the deepfacelabab NVIDIA RTX3000 series and don¨t know what option for train my extracted face, thkas for the video
You can learn hair with it if you do a full head model. You can include things like beards, the person's neck, and hats/headgear if they are present in the source material used to create the model. Clothes like shirts/pants/etc cannot be added because of the way the software extracts the faces from the original images. Hard to explain if you've never used Deepfacelab to create a model, but yeah, the head is the target area.
Can u make video convert from celebrity setfaces into DMF file? Im sure many user confuse 😂 because after download model then extract they just showing aligned only pictures. I hope u can make tutorial ❤ new sub from indonesia
I don't understand what you're asking. If you're asking how to make your own DFM from celebrity images, I have a few tutorials on my channel that cover how to do that.
Update: I've noticed several comments of people asking where to get access to DFM model files to use with Deepface Live. I make free models regularly, which you can access here:
drive.google.com/drive/folders/1tH-8SXb5R4-W1c3wIVXy4aYnZOP0Om-1?usp=sharing
------------------------
I apologize in advance for the weird sounding voice audio. Apparently OBS was recording both my XLR mic along with the mic built into the Logitech Brio.. the result is a bit.. unsettling, lol.
*Update*
I found a walkthrough from the developer (Iperov) about what settings to use to create your own 224 res model that you can import into DeepFaceLive:
github.com/iperov/DeepFaceLive/blob/master/doc/user_faq/user_faq.md
Maybe do the scene where Neo and Trinity unite, but its Jensen with the Wang exploding facial expression and Lisa.
Thats cool. I expected some latency. Sure you dont want to step up your shock value? The potential of this technology is really interesting. Cant wait until you upgrade.
Can u make Cristiano Ronaldo model....
You can make pleace Putin?
I want Kris Wu... Can help do one?
You're the only video that shows how to actually use this not just install it. Thanks.
I have a tutorial on here that show how to make your own models if you're interested.
@@DruuzilTechGamescan you train my model if i send you 10 pictures?
Please which camera are you using? My FPS drop to 1FPS and I don't understand it.
At 11:45 which version of deepface lab is that? I don't have "6) export SAEHD as DFM" I already trained a bunch of models with Quick96. Do I have to start over and train with SAEHD so I can export DFMs?
Yes, you'd have to use and SAEHD model to export as DFM, so a Quick96 will not work. That said, you should have that batch file in your folder if you're using any of the last couple versions of Deepface Lab.
@@DruuzilTechGames cool thanks for the info
@@BeardofBeesPool Tested my Arnold model. It looks "ok" but it's not as good as I thought it'd be. I'm kinda curious how much they trained the models that come with DeepfaceLive.
Hey druuzil, i have a question, is that normal that i got very low fps when i use deepfacelive + playing a game ? i went from 30/50 fps with dfl alone, then 8 fps when playing + recording a game + dfl
lol yeah. Deepface Live is extremely system intensive.
@@DruuzilTechGames i'm surprised cause i have a i-9 12900k + 3080, so i guess there is no way to make it work :(
Do your trained models work for regular deepfakes as well or just live?
There seems to be some debate on that. There's an old version of DF Lab where you can supposedly use a DFM file (a model exported into the format that works in DF Live) and put that back into regular DF Lab, then apply it to a video clip like you would normally do with DF Lab. The problem is you can't train the model any further that way, and usually when you use a new video clip (like if you want to put Tom Cruise into an episode of "Friends" for example), it may not work that well because usually you'd have to train your model against that new video clip for a while before the final merger process.
I hope that made sense. Anyway, I have yet to see someone do it, but someone was claiming in Discord earlier this morning that there was a process to do it, so I plan to test it out at some point. If it works, I may do a tutorial on it.. but generally, making a model of your own would offer superior results for a few technical reasons.
Thanks. I am wondering though, will this work for all websites that use Webcam? Will it always show the live deepfake face of me, not the real me face?
I don't know, I haven't tested it outside of the DF Live environment. When I livestream it via OBS, it's just OBS capturing the screen and sending it to UA-cam. I'm not sure how a chat website or something similar that uses the webcam would be able to interface with DF Live.. Something I'd have to play with.
@@DruuzilTechGames Can you try it out for me please? I don't want to download it if it doesn't work..
Nice video. Just to add to it cos you kept mentioning resolution and I noticed yours was in default setting. I normally increase mine to 960 and it’s fairly clear and detailed all round
Resolution caps out in DF Live at 640 for WF, 768 for Head, and there is no hardware that can run models at that resolution with a decent frame rate.
I can't use the Nvidia gpu on main processes... Any idea how to be able to see my gpu in every option?
Laptop or desktop?
I just need to ask that can I select my own model ? like my friend or some other person who is not common. because there is no any choose file option for getting my own known ones photo saved in my laptop. I want to deface other person photo. please help me
You can create your own models if you have proper hardware and are willing to put in the time/effort. I have tutorials on this channel that go over the process. It is somewhat complicated though and as mentioned, time consuming.
How are you liking the WX6800? planning on getting myself one as im currently on a Titan V and its doing fine but i would like some more umph and id be able to get the W6800 for a nice price if i can sell off this Titan GPU.
I'm having problems with it actually. Trying to do an RMA with AMD to get a replacement. Not 100% sure if it's the card, AMD's drivers, or Deepface Lab to blame though. Trying to get another AMD card locally so I can test and see if the problem is AMD in general or not.
This could actually be very useful in corporate environments. Imagine people working from home in a zoom call, unshaved , or without make up ( women of course) can now enter a zoom call with their office face. There would be a market for it I guess. Or UA-cam presenters that want to look a bit different or better online can rework their face digitally and iron out the imperfections for their online presence
Yeah I could see that. Or live interviews etc for celebrities who want to look their best but don't want to have a makeup artist spend time/$ on making them look good. Could probably be used in a lot of ways.
My image quality is blurry and not sharp, can you give me some guidance?
Are you the person I talked to on Facebook?
Hey thanks for this I've always wondered what Kyle Dunnigan uses for his videos he is pretty big at doing this do you have any guesses?
If you watch his videos (I'm a big fan of his too), the models are pretty bad quality and don't really look to me like they came from DeepFaceLive.. Maybe FaceSwap?
Did you watch his video "Star Boat"? Hilarious.
@@DruuzilTechGames Yeah even the bad quality just a spot on accent is doing wonders, he will do his wife as a baby and it's near perfect
@@MagnaRads I tried messaging him a few times on his channel to see if he'd let me help improve the Deepfakes but didn't get a response. I think in fairness the bad quality actually makes it funnier anyway.
@@DruuzilTechGames Yeah I've tried asking many times what face swap software he is using that works with teams
I'm a big fan of Kyle Dunnigan and I heard he uses Face Swap Live.
hi i have developed deep fake buy there was no lip moment from model when we are talking. Can you suggest what i need to do
I can only assume you didn't use enough src material that has the person talking, or it's an issue of how you trained the model. Did you follow any of the tutorials on my channel?
Bruh , good video but what apps does this work on ? And does it work on a 4gigram laptop ?
As far as "using it on other apps" goes, I think that pretty much requires using something like OBS and a virtual camera to "clone" the DF Live output into the app you're trying to use it with. I've had limited success with that though and haven't screwed with it much in quite a while. I tried using it with Omegle but wasn't able to get it working.
As for what kind of hardware you need to run it, you need a dedicated GPU (if your laptop only has 4GB of system memory I guarantee it's not fast enough to run this).. I think the recommended specs are something like an RTX 2080 Ti (or RTX 3070 which is equivalent) or faster. That'd be to run 224 resolution models.. 320 resolution (which of course looks better) requires even faster hardware.
It's not a filter, it's literally deepfaking you in real time, so it's unfortunately quite system intensive.
If we have lower specification laptops. In that case what we need to do. Can we use cloud gpu providers like(papersspace).
Any suggestions will be help full.
I don't see how you'd use a cloud service to run DF Live (assuming that's what you meant).
@@DruuzilTechGames Yes i have meant the same. Is that possible to do?
If yes suggest me how to do
If no why it isn't pssible.
Because it's in the cloud..? You can't run a local application like DF Live in the cloud and interface with your local webcam without extreme latency I'd imagine.
Sorry for more questions
But there are people who are playing games using cloud gps. Is there is any delay how they are able to play games
@@RavulapalliNareendra There is delay, yes. It's like a remote desktop session where the game is being played on the server and you just see it on your screen.
Great video!!! I love how clear you make these explainations. I've got one question though. How can I create my own model for deepface live and how can I add it to deepface live? Okay that's 2 questions. I apologize
ua-cam.com/video/_bc3SPbCdW8/v-deo.html
Enjoy.
I am having problem swapping the face even after I select a picture I want to swap. It’s only source frame and aligned face . It’s not showing face swapper or final frame. The deep face live output isn’t showing as well. Pls can you help me out.
Did you select the merged face option from the bottom right
pls where can i download DeepFace Live setup the links you posted are just for the the Graphic drivers not the deepface app ..... HELP
What are you talking about. There are no graphics driver links in the description. DF Live is definitely still available at least on the MEGA download link.
Hello, why i have no sound?....i record a video with obs, then i deepfacelive it but there is no sound....at the window out, not even when i record with obs....
why?
thnnks!!!
Deepfacelive outputs just image files. You'd have to edit the audio back in with some kind of video software like Adobe Premiere.
@@DruuzilTechGames ahhhhhhh thank you !!!
Awesome video man! Thank you for this and God bless! :)
Thanks!
Pls, i want to ask questions. Why Dell G5 Geforce Gtx 1660ti 16Gig Ram not running the software easily
Because the 1660 Ti is old and slow. Deepfaking requires fast, modern hardware.
Will Dell G5 of 32GB Ram run the software easily
The GPU is what matters mostly, so no. The 1660 Ti is slow and old.
It run it smoothly, am using Dell G5 and it's perfect
Can I upload another picture on the software apart from the one that was given for us to use ??
if you're asking about using other models than the ones that come with DF Live, yes you can do that. I have a full tutorial on my channel on how to make your own, and I supply free models all the time. Check any of my newer livestreams for a link to my Google drive to download free models.
I successfully installed deepfake but how do I launch it after rebooting my laptop
There's nothing to install, it's just a folder with some files in it. Run the batch file in the main folder.
my deepfake is only showing cpu under yolov5 what should i do i have RTX 2070 with maxQ
Disable the integrated GPU in the device manager so only the dedicated 2070 is available to the system.
Can I able to use it on discord ? Cuz like I can't use dual software of webcam
I don't know. I think you can use OBS and Virtual Camera to set stuff like that up but idk how.. I've had very poor luck with Virtual Cam.
@@DruuzilTechGames thank you for your quick reply. It really helped alot 🙏
Please how do I set up the obs to connect and stream from deep live I want to use it I'm skype
I don't know how to connect it with Skype. To use OBS for DF Live I just do a "Window" and tell it to use the Python.exe DF Live output window.
what do you charge for a 512 model of Tom Selleck?
512 really isn't viable in Deepface Live. Unless you have an RTX 4090 video card, it basically won't run at all, and even if you do have a 4090 the frame rate is pretty bad. 320 is much more realistic.
Do you know what video card you have?
@@DruuzilTechGames i am working on Tom Selleck 320 head model right now but i foooked up so i had to restarted it all lol.
@@DruuzilTechGames i made a 512 model of my wife when she was younger :) it brought so much memory back for us with tears flowing freely of course on her side.
@@DruuzilTechGames have you made any models with full thick beard?
@@DruuzilTechGames btw bro thank you so much for all your DFM models, so nice of you to share with the public.
Hello deepface does not list my GPU under face swap DFM only says CPU even tho i turned off my processors gpu any help perhaps?
Did you reboot after disabling the IGP
yes for some options it still only list cpu just like before @@DruuzilTechGames
I don't know then.
bro do you know,why my deepfake has a delay.it is soo big delay
Can you elaborate a bit? Did you lower the delay value in the bottom right setting to like 5ms?
Can I use 2g dedicated graphics for this?
Eh.. maybe, but on extremely low settings.
Hey deepfake live doesn't detect my gpu for some of the options such face detection and face maker where as few detect my GPU...i have a 4gb RTX 3050
I assume you're using a laptop. Shut off your integrated GPU so it will be forced to use the dedicated 3050.
Incredible tech. Can't wait to see the funny stuff this is gonna be used for in the future.
Fun but dangerous in the wrong hands. :D
Why the face detector can't stabilize. It's jittering? it makes the end result of merging face, also jitter.
DF Live is not intended for creating videos, just livestreaming (per the developer). It doesn't produce as good of results as DF Lab.
Please I need help in creating a dfm file
i have an alienware m16 will it pick the software?
Need more info. What are the specs. Also what does "pick" the software mean?
How can I do produce a model dfm file
Is there an application that can create a dfm file
I have a tutorial on my channel that covers the entire process. Check it out. :)
Please drop the link I can’t find it because your videos are much
ua-cam.com/video/_bc3SPbCdW8/v-deo.html
Can rtx 3060 or 3070 run this on laptop.. Dedicated gpu 8gb
Run it, yes. Not sure what framerates you can expect.
Brother thank you so much for you video. Could you please guide me on the part where I actually wish to use this in a live video conference format. For example, in Discord. How do I export the DFL picture into a different program?
It involves using OBS and virtual camera, but I've not used it for that personally. You'd probably have to ask in the DeepX discord channel.. there's a couple guys in there that know how to do it, I just haven't had much luck with the virtual camera stuff.
Great video and explanation.. have to try this out once..
Thanks!
How can i also see the GPU and not only the GPU under yolov5? 😢 I see the GPU only under Face Merger...
No need to spam. Is it a laptop or a desktop computer?
This is awesome!!!
Thanks. It's pretty cool tech.
Can you share how to bust someone using this software? This is obviously popular among scammers, does the face de-sync if it's obstructed? What needs to be covered? Half the face? The eyes? It's obvious if the face is turned to the side that it will get distorted, what other ways can you break this software?
It definitely doesn't handle facial obstructions well, unless the model is specifically trained to help with that which is time consuming and most people don't bother to do it. Unless the person does something specific during the video to destroy the illusion however, idk. PIxelation around the eyes, inside the mouth... facial distortion at certain angles if the model didn't get properly trained or the faceset was insufficient to cover usage at all angles.
hi i just tried it, but how i transfer that merged image to obs for ex ?
thanks!
In DF Live, use the "Merged Output" window (so you can see yourself as the deepfake character in a larger window), then in OBS do "Window Share" and select that Merged Output as the window you want to share.
@@DruuzilTechGames
thnks !! the window thing i cliked but did nothing... ill try the source choose the merged window as the source for obs (if the bigger windows still doesnt appear) ill try later thanks for the quick answer !
Whats the process to create a deepfakelive character vs regular deepfake?
It's pretty similar. If you go to the links I provided in the description of the video, you can find a tutorial on the creation of a DF Live model.
Basically you still make the model in DF Lab, at 224 resolution. You need a good source faceset of varied images (this is the character you want to see yourself as in the live broadcast). Needs various lighting conditions and facial expressions, looking in all directions. You then train that set of faces against the "RTM" faceset (found via the "Magnet" link on the Github site, again link in the description) using the instructions for creating a Deepface Live model (again, linked). Once you've trained it sufficiently (mine started to get decent around 1.6-2m iterations), you export a "DFM" file and place that into the DF Live models folder, after which you can select and use that model in DF Live.
I’m getting an error message when I try to launch it it says unable to load openCL library
IDK, I've never had any issues launching the application. What GPU do you have? Are you using the correct version for your GPU?
@@DruuzilTechGames I a core i7 720q and an nvidia gt 330m installed dircectx 12 when I check on my nvidia panel it says directx 11.0 but on dxdiag it says directx 12 so I taught it should be able to run the directx 12 version but I get the openCL library error when the nvidia driver is not installed but now that’s it installed I’m getting a different error
@@DruuzilTechGames AttributeError: function ‘cLEnqueueFiLLBuffer’ not found
Please what’s the name of the application and how do I get the download link
It's Deepface Live and the link is in the video description.
can i use it microsoft team and google meet for video call it will work?
I think you have to use a virtual camera with OBS to set that up. I haven't personally done it.
hello, can you share new link for basic the software. Because the old link already not found. i want to try to make new file dfm... Thankyou
What link doesn't work? Both links in the description are working.
Can i upload a picture of my own and use it to do a deep fake?
Not sure what you mean by uploading a picture. Upload to what? Deepfakes require thousands of images, and a fairly long period of training time to get the model to work properly.
I have tutorials on my channel that cover how to do it if you're interested in learning.
Is this available for Google collab?
DF Live runs locally on your PC. I don't think you could run it in a cloud environment, and there'd be a ton of latency if you could. Additionally, Colab banned using Deepface Lab, so I doubt they'd support DF Live either.
congratulations nice video I wanted to ask you how do I export a video? a thousand thanks
You can't. You can only export the individual images, which you then would have to import into something like Adobe Premiere Pro (as an "image sequence") to convert it back into an mp4. It loses the audio too, so you'll need to pair up the audio track with that in Premiere to export the final product.
@@DruuzilTechGamesthank you very much pity it would have been nice to export a video
Well you can do that if you want to make your own models using Deepfacelab. Better results too than DF Live. Just much more complicated and time consuming.
Can you give me a link to the training tool of the dfm model? Thank you so much!
mega.nz/folder/Po0nGQrA#dbbttiNWojCt8jzD4xYaPw
@@DruuzilTechGames Thank you very much for your reply, I'm still learning how to use it.
Hi how do i Import DFM files ? please help
Download the DFM files, then go into the Deepface Live folder, then the models folder and put the DFM files there. Then you can select them while running the DF Live application.
Can create new models that are not in the list other than Keanu reeves, Arnold Schwarzenegger etc?
Yes. I have a tutorial on my channel that covers how to do it. You can also download the ones in my Google drive which should be in the video description.
Is this possible on Mac?
No, Mac isn't supported
Will it work with the Rtx 370 Ti
Yes.
Thank you very much for sharing the model
You are welcome!
I am using it on rtx4060 8gb but it is lagging so much placing the udp ip camera in split camera
I've never messed with the IP camera, can't advise you but the RTX 4060 is not a fast card. Even on a 4090 the performance in DF Live isn't amazing.
Is there a way we can download more faces for Deepfacelive?
Check the video description. I have a link to my Google drive in there which has about 50 models you can play with.
I don’t mean to bother you. But have you made a video on how we would upload the DFM in your drive to our Deepfacelive
I mean how do I upload the DFM in my Deepfacelive 🙏
There is no "uploading" to DF Live. You just download any of the DFM files in the drive, and put them in the DFM folder inside the DF Live main folder. Done.
And how do i get the deepfake software
Please check out some of my newer videos on the subject. I did a full tutorial on how to create a model for Deepface Live. DF Live and DF Lab are 2 separate software suites. You create the model in DF Lab, and export what's called a "DFM" file, which you can then use in DF Live.
didnt understand your section of creating a model of your cchoosing
That's an OBS limitation.. it shows the window but not what's being selected.
Please I can't use my gpu file in the deepface camera
Is it a laptop or desktop you're using
Hello… I’m running on 8GB rtx 2080 laptop It won’t give me good results. I’ve seen people run it on 2070 and it works well mine is shaky and looks very unreal
Has nothing to do with the GPU, and likely something to do with how you're training the model.. Like what settings, when you're enabling them, for how long, etc.
*edit*
Are you referring to how well the models that come with DF Live are running/look on your system vs his? Are you using the same models the other person did? If not, it can vary from model to model how well they perform based on how they were trained.
@@DruuzilTechGames yes… I’m referring to how well the DFM are looking. I didn’t train them myself.. I Download one out of yours and it was a disaster. Do you think it has to do with my pc’s camera?
@@DruuzilTechGames it won’t even run the models that came with DFL very good. Do you have discord so I can share pictures there
It has to be some kind of settings, and maybe the lighting in your room. Soft, white light tends to work best. You can find me on Discord at "Druuzil_TG" if you want to send a pic.
Couldn’t add you in discord…
Can you drop your Discord links?
This is so cool.. is there a way to slim your face?
Maybe with specific camera lenses? Not from within the software itself as it currently exits.
@@DruuzilTechGames ok thank you
Hello I'm from Indonesian, I'm shock to see this aplication live deepfake really creazy and awesome 😆👍
Check out some of my more recent livestreams. The models have gotten a lot better since I made this video. :)
can you make a video how to install this on a mac no one has the install process for deepfacelive they just show what they do with it ?????????????
I don't believe it's supported on Mac. I know Deepface Lab isn't.
Is this the deepfake software or a livestream app
how i open this its only winrar cant open finish download thanks
It's an exe file. You don't need anything special to open it. Just double click on it and it should work.
Is there a way to extend the facemask to the neck legion
Yeah but you have to create the model that way in the first place, and I'm pretty sure you can only do it with Head models, not "Whole Face" models. Even then, it doesn't work super well unless you remain fairly still/don't move around.
But I noticed it on your Morgan Freeman model how did you do that
It's a head model.
does this deep fake of a thing contain a voice changer in it!?
Nope.
bro im having problem with mega links can u provid me gdrive link
I'm not aware of a GDRIVE link for DF Live. What problem are you having the with the MEGA link? It works fine for me.
Can this work on windows 7
I have no idea, sorry.
why is mine lagging? i have nvidia and all that... is it my camera?
What system hardware are you using? Did you change the delay to 50ms?
@@DruuzilTechGames im using windows 10 for gaming acer, nvidia gtx 1650 Ti...
@@DruuzilTechGames i have intel too...
@@aam705 Is that a laptop? I don't think the 1650 Ti is a desktop card. Anyway, that's a pretty low spec machine to be even attempting something like this. To give you an idea of the systems I'm using and still only getting ~30 FPS at 480p, I have an i9 11900k/RTX 3090 in one system, and a Threadripper 3960x/AMD W6800 Pro in the other. Both get similar results.
@@DruuzilTechGames yes its a laptop.. how do i solve this then? i'm pretty new to this kind of stuff.. but i wud love to try
I really can't change the skin tone or colour... The models looks different with different skin colours
Are you talking about with the DF Live settings? The only CT mode available is RCT, but it should help.
Yew the deepface life skin colours can change, my original skin colour is black and the output is still black
well it's blending the mask into your face. If you don't want it to be black, shut off ct mode.
Does this work on Mac book
No.
Nice job but can voices also be changed?
Real time voice changing isn't super good yet. There's like 1 or 2 applications out there but they're not very good quality and there's quite a bit of lag because it has to change the voice on the fly.
how to change my own model?
Are you asking how to make a model?
Can we train models on the deepfacelive if yes please give us a tutorial on that🙏🏿
I have 2 large tutorials on my channel that tell you how to do it.
Don't train any model, just get New version of deep face live and you can use any pictures of your choice
Awesome video. Can I install it on mac?
Unfortunately MAC is not supported by the developer.
Okay! I windows. I say the link on your UA-cam Channel. How can I set it up after downloading?
@@fidzack1027 I have a tutorial video on my channel that covers use of Deepfacelab, please give it a look and if you need help let me know
Hello, please my "swapped face" keeps flickering
Do you mean flickering..? Is this a model you trained yourself?
@@DruuzilTechGames yes flickering... I just got the PC Alienware x15 r2 3070ti 8gbvram...
I tested it with a dfm(Keanu reeves) downloaded on the dflive.
Would I be able to train a model with the PC?
idk why people keep buying Alienware laptops to make deepfakes, lol. I've seen so many people buy similar systems. Laptops are not a good idea for training models 24/7. They perform poorly vs desktops and get very hot.
Anyways, the 8GB of VRAM is going to limit you. You "may" be able to train the 224 RTT model, but probably more like 192. Not ideal.
Thank you, I was jusy about to get the Alienware rtx 3090.
What should I get sir,
what better laptop? and spec?
please which laptop should we get?
Don't get a laptop, get a desktop. An RTX 3090 is a fine card, you can do up to 512 resolution models with 24GB VRAM.
It's not working on my pc
Window 10 pro , 4gb
4GB of RAM or VRAM? What video card?
Hey, what if you're turning into George Costanza? "George is getting upset". : )
Hah yeah that'd be a good one for me. I'm gonna see how well the Arnold model I'm training ends up working and then I may do that. So many Deepfakes.. so few computers...
Thanks so much for this.. Please do you mind doing dfm of people that are not so popular.. Like USA soidiers. I think it's going to be more fun 😊
Non-celebs are not in the public eye and are not viable without express permission. Besides, part of making UA-cam videos is so they'll get clicks. If nobody knows who the person is, nobody is going to watch it.
@@DruuzilTechGames i know not regular people.. People that are not super famous in the States.. Or just any celebrity military active personnel..
thank for this video.. I was thinking how u can do it at live it is possible.. now I know what to look if I think ppl I am talking on live is not real
I have a couple tutorials that show how to make your own model if you're into that sort of thing.
thanks this helpt alot
What's in rtm wf face set...and how to use it
Please have a look at the tutorial videos I've made on how to make a Deepfake. Basically though the RTM faceset is about 63k images in a pak file that you can train your model against so it learns a ton of faces, lighting conditions, angles, etc.
@@DruuzilTechGames I can't understand..is that pak file have built in actor/actress images
@@jaya_surya_a5118 Are you trying to make a Deepface Live model
@@DruuzilTechGames no ..I don't have source image...idk whats in that..the pak file is 8 gb..can I use pak file for regular deepfake video thx
@@jaya_surya_a5118 If you don't have source images you can't make anything. The RTM faceset cannot be used to make a regular video. Please go checkout the 2 tutorial videos I made on my channel which explain 100% how to make a Deepfake from scratch. One explains how to do a regular model, one explains how to make a live model.
Can it lipsync other audio live
It just swaps the face. has nothing to do with audio.
@@DruuzilTechGames I guess I'll have to figure that out on my own lol
I guess it depends on what you're trying to do. If you want to livestream with DF Live, use OBS (and OBS will allow you to select an audio source). If you're trying to apply a deepfake to a video clip using DF Live, then it doesn't add audio and you'll have to use some kind of video editor like Premiere Pro to add an audio track back in after the fact.
do you accept crypto donations?
You take your time. Love your work
No but I do have a PayPal if you want to go that route. It's linked in most of my videos.
Hi, great video, I can´t find the way to train my own face, I extraxt the face but where have to train it ? I have installed the deepfacelabab NVIDIA RTX3000 series and don¨t know what option for train my extracted face, thkas for the video
There's a link to a walkthrough in the pinned comment.
You won't be able to use any face of your choice cos you're using old version of Deep face live, Get new version and you can make use of any picture
can it work on AMD
Yes.
Please teach us how to do live stream with deepfake :(
Do you know how to use OBS?
@@DruuzilTechGames yees
@@DruuzilTechGames i want to use it on what'sAPP for pc lol
I've never used WhatsAPP on PC. I use OBS to broadcast the DF Live application window to UA-cam.
@@DruuzilTechGames i used to be able to do it on youtube live but i forgot how lol i am an elderly guy
Why doesn't it change your clothes and your hairstyle?
Not sure if this is a serious question.
You can learn hair with it if you do a full head model. You can include things like beards, the person's neck, and hats/headgear if they are present in the source material used to create the model. Clothes like shirts/pants/etc cannot be added because of the way the software extracts the faces from the original images. Hard to explain if you've never used Deepfacelab to create a model, but yeah, the head is the target area.
I'm sure you can find software that does filters etc to add clothes, I'm just not familiar with any, and they wouldn't be a deepfake, just a filter.
@@DruuzilTechGames thank you so much, I find this information very useful and helpful 🙏
Hey can you give me a private tutorial please , im having issues
What issues are you having.
Can u make video convert from celebrity setfaces into DMF file? Im sure many user confuse 😂 because after download model then extract they just showing aligned only pictures. I hope u can make tutorial ❤ new sub from indonesia
I don't understand what you're asking. If you're asking how to make your own DFM from celebrity images, I have a few tutorials on my channel that cover how to do that.
@@DruuzilTechGames which one mr?
Can you make a video on how to use photo instead of model
idk what you mean.
@@DruuzilTechGames I mean like animate Joe Biden face. Photo animation
I do not have the map called ''workspace''
What do you mean "Map"
Hi thank you for the video tutorial on the deepfake live how do you change voice
Other software. I didn't use any in this video, I just do my own voice impersonations.
@@DruuzilTechGames please what software is best for the voice?
This is scary
Yup, in the wrong hands... for sure.