After you watch the video, I've got some extra tips that I missed in the video in the description! So, make sure to check those out for even better swaps!
For better face consistency make sure you've got the different angles selected. You can also enable 'many' on the face detectors although this will slow down the processing. Another issue you may have had is playing with the face distance slider.
Totally welcome! And I get it-- but here's a funny thought for you: Soooo much media consumption is on mobile these days-- and, this kind of stuck me the other day as I was watching one of my own videos: On mobile, a LOT of those imperfections kind of vanish. I know most of us look at all of this with such a keen eye for detail-- but, the larger world (sadly, maybe?) really doesn't see it.
FaceFusion is such an annoying mix of greatness and noob programming. Its AI algorithm/processing: Greatness. Its code: Noob. It writes every single frame to disk as PNG and processes them step by step, rewriting the data over and over, wasting your hard disk's durability. It's an SSD wrecker. For a 1080p video this can easily mean 400 gigabytes of data written to the disk. It makes me not want to use it. The proper design would have been to read 1 frame at a time, process it in-memory, and send that frame to the video encoder live, without ever landing on the hard disk in-between.
If you'd like to improve it they have a active discord, the 'early adopters' section is for FaceFusion Next with experimental features. FYI I have no idea how your using 400GB of space unless its a v long 1080p video with enhancements? Maybe look at some of the other more in depth tutorials if you've not done so already
Cool! It works really well! I always recommend that people use a separate computer to run any of this AI stuff. It's the best way to avoid exfiltration of personal information and password stealing Trojans. Remote control and unauthorized processing (i.e. bitcoin) are still potential issues, but since it's on separate computer, it's something you can monitor and deal with in the future.
13:35 MOST IMPORTANT PART Thanks for the video Tim. 👍I see a handful of gems in an ocean full of horrible misuse. Shame. How will I convince my wife I'm _not_ cheating on her! 😆
suppose there are multiple people in the video and we want to change various faces then it works or not? Similarly, if there is 2 or more people in the frame then how does lip sync works?
You should have targeting options (right person, left person) on the right had side. At the bottom you’ll see a pull down that says “instant run” or something like that. You can use that to schedule “jobs” where FF3 will run once w/ one person, and then again with the other. TBH, I didn’t have time to play around with jobs for this video, so try it out and let me know how it works!
Very interesting. Thanks a lot dear Tim. I try to install but the installation stops with an error when trying to install Microsoft Visual Studio. Anyone has an idea how to fix. Thanks in advance.
Best bet here is to hop over to their Discord. Lots of very smart folks over there who can help. If you put in a support ticket, they're really good about helping out too. It's funny, as a community driven project, they're actually better at support than most companies are!
It seems that you can't install it if your windows user name has a space in it. Mine is my first and last name and it fails to find my user folder because it is only looking for my first name.
You should be able to point it somewhere else. Check in the settings. I have it installed on an external drive. Granted, that external drive is named T7 (the Samsung portable SSD that everyone seems to have) and it installed fine.
Roop is great too! In my opinion, I lean to FF, but do try it yourself to see what you think. It’s free! (Plus, tons of other stuff on Pinokio if you aren’t already on it)
I'm on a Mac. I know, one of the few in the AI space that runs mac. I just have a (now older) M1 Macbook Pro. But, running fine. To be honest: Pinokio and FF3 will run on about anything. Even a Dell Potato. It'll just take a lot longer.
How is this free? It looks 100% free with no small prints, so what gives? (Also... there is no standard size for croissants. As long as it is shaped as a crescent moon, it's all good).
the slider "face selector age" is not used to tell the AI the age of the person but it helps select people on the picture according to their age (like gender/race/distance selectors)
I actually have it running on an M1. Did you run the patch? If you’re having issues, best I can say is open a ticket on their discord support. They’re really good about helping out. Again: this isn’t a company, it’s a community…which means the people actually care and want to help!
Yeah, the take on this was more along the lines of "outside of Comfy" and more directed at the intermediate or starter level. People apparently take their faceswappers pretty seriously! haha- I updated the title and thumbnail and removed "Best"-- Hey, at least I didn't say "Everything Just Changed!"
Thank you thank you! It's so good! And hopefully, reaching out to our Peanut friend, some more folks will join us-- at least, those who were on the fence w/ that WinDefender warning!
@@TheoreticallyMedia Yeah, it's good that you explained that 🤓. Hopefully raising awareness on the reasoning behind it will allow "Mr. Peanut" to stick to his original guns, too 😅
So, I didn't get into it for this video-- namely because I didn't have a lot of time to experiment with it, but at the bottom of FaceFusion, you'll see something that says "instant run" or something of the sort. Sorry, not at my machine... Anyhow, that's a pulldown where you can use "Jobs"-- this will allow you to run FF on one character, then after it is done, it'll go back and do the other. Again-- didn't have time to dig into this feature, but it is there!
That depends on what graphics card you have. The CUDA option is there if you have an Nvidia card. Don’t worry about it though, those options just act as a speed boost by using the GPU. It’ll still work.
I'm getting an error while installing the "cloudflared" part of the facefusion installation, and I just can't get any further. Do you know anything about this? The error message goes by so quickly so I'm not able to read what the issue is...
AI Face Swap IO is free, done in browser, does not require an install. None of you should be installing random software you find on the Internet. Just use the Web stuff.
@@TheoreticallyMedia In my opinion, inviting your followers to download possible spyware off Internet randos is going to screw you over eventually. "I spoke to the owners and they seemed like swell guys" is not the kind of antivirus you think it is.
@@dirtydevotee Most software, whether open source or closed source, has bugs and exploits. Secure software is exceedingly rare these days. Even your hardware can have exploits and malware running under the level of the operating system, nearly impossible to remove.
@@ILoveTeles I'm on Linux. So if it's in the Linux Mint "software manager" on my taskbar, its allowed on my computer. It's 2024. There is literally nothing else that you can't find on the Web. Photopea, Dezgo, Krea, six video providers, and hundreds of everything you can get on Git. What are you people doing risking bricking your machines for a cheap thrill?
I've been using AIVideoFaceSwap for some time and am really happy with it. The latest update which allows a quality setting of 1.5 (previously 1.0 was the max) also handles head movements and characters making faces very well. Only disadvantage is the processing time, a 4 minute video took a good 4-6 hours on my 3080 at the highest quality setting.
It can be. It'll take anything (within reason) and swap. Like, people have asked about swapping animal faces...that....while it'll do it, looks horrifying. So-- within reason is what I'd say.
You're so funny! Btw, I have a question about a face leaving or entering the frame, and I wonder if you can advise me on how to avoid "glitches"? Or do you have a video that talks about it?
Installed it manually now, without Pinokio. Works, but I always get totally blurry output, no matter how crisp the input images are. Can someone here address this? Quite sure that I am not the only one with this issue...
Sorry Pinokio is a No-Go - it doesn't notice or respect already installed conda and git versions and it even creates trouble if you want to uninstall and reinstall it after it stops to work. Which it did on my system for no apparent reason after a few month of not using it.
I'll say, that's one aspect of Pinokio that tends to be a bit wonky-- the whole uninstall/reinstall thing w/ previous version. I tend not to jump on any updates w/ any software-- but, w/ Pinokio, I tend to wait a few days or longer for bug quashing. But, at the same time, it's an open source project, and not commercial or anything-- so a lot of leeway. If you're having issues, high recommend you go let them know in the Discord. They can't fix it, if they don't know about it!
@@TheoreticallyMedia I've got a massive bunch of AI tools installed manually (no problems being a computer scientist) and also use 'Stability Matrix' as a launcher of my main AI GUI like A1111, Fooocus etc. I guess it makes more sense if I stay on this path instead of risking Pinokio messing with my perfectly running stuff.
I was hoping you'd show if it works in different media styles. On the extreme end of things, can it face swap /fuse from a real photo to a cartoon / anime? I remember a Corridor Crew video taking faces and putting into different art styles, like Yoji Shinkawa (Metal Gear Solid art). But that was about two years ago (video title: "Is this the death of VFX").
I remember that one! I’ll have to do some test, but I don’t think that this is where FF’s strength is. You’d likely end up with something that looks a little cursed. It’s more meant for realism. But, I’ll give it a shot! Why not?!
@@TheoreticallyMedia I, and probably the rest of your viewers/subscribers, appreciate that you do all the AI platform testing, and respond to comments so efficiently. It's one of those things I would like to have more time to do but thankfully you're here. Thank you.
@@TheoreticallyMedia I tested this... Animated cartoon video of a real person made using Flux + KLING and the FF3 for swaping the cartoon face with real face photo of the person... Frankly it worked out amazingly well and looks gorgeous... The cartoon textures and lighting etc are very accurately preserved and the real face blends beautifully.. making the whole thing look like a very high end studio cartoon movie
That dev either lied to you or you didn't understand it and communicate it bad but getting rid of that warning has nothing to do with joining any "wall garden", He could pick any and I mean any (not related to Microsoft) legitimate code signing certificate from a trusted Certificate Authority (CA) like DigiCert, GlobalSign, Sectigo, etc., and that will make the software trusted by Windows and avoid the security warnings.
Great overview, I’ve been using insight face swapper, since you featured it and this a big leap forward for me, even the facial expression palette…I just have to get over the install, but thanks for the install simplification and general starting point settings. Much thanks Tim. PS. AI audio has been sleeping or I’m late to the party , I hope we have this much control with real time audio mangling and transforming , although vocaflex has been doing a great job of sound imprinting to create new vocal voices, but I would like to see more sound synthesis that really can handle morphing sound in real time , kinda like realtime synthesis with a synth plant engine. 1nce again cheers Tim.
Thanks Tim. Do you use comfy ui much after finding tools like Face Fusion 3, or do you find that Comfy is still the ai filmmaking tool with the most control?
Does anyone have a suggestion for a good AI platform to match green screened actors to background footage, lighting , colour grade etc? Viggle and wonder dynamics do a great job with CGI characters. This actually does it with faces. I would think matching footage of actors would be easier?
Oh, I think you can make money with it in certain cases: like, if you’ve got your own film you’re using it for…like, a remake of Face Off (haha, the first thing that came to mind), and you’re using your own actors. But yeah, I get your point!
@@TheoreticallyMediathe way I understood it when I looked into FF, a few versions ago, was that the output is very strictly non commercial. Only the code is under the MIT license. Or something like that, I’m (clearly) not a lawyer. Which is a bit of a deal breaker if like you say you are an independent or hobbyist filmmaker. One day maybe you get the chance to show your short film as part of a compilation on TV or somewhere but then you can’t because one of the tools you used has this limitation. This is also the problem with a lot of other tools, like ComfyUI nodes that you only use as a small part of your workflow. Yet if you are to follow the rules they dictate your entire outcome.
Where we are headed with A.I, in general, is nothing short of amazing. Can't wait to see where this goes in the next 10 years. Keep up the great videos.
You're a good man for getting that Install messaged cleared up by the developer Tim! 😎 Figured that might have been the case when I installed it a while back, but good to get confirmation anyway :)
Haha, yeah- I get hit with that everytime I mention Pinokio in a video, so I decided to finally just get the word out-- so, at least next time I cover it and someone asks, I can just link to this video!
I can't get Pinokio to run correctly. It keeps trying to install stuff and saying Error: ENOENT: no such file or directory, copyfile 'C:\pinokio\bin\miniconda\python.exe' -> 'C:\pinokio\bin\miniconda\python3.exe' I have both miniconda and python running so I don't understand what's happening.
Hmmm-- might need to fully uninstall and then reinstall. That sort of thing has happened to me in the past. To be honest: Best bet is to check in on the Discord. Pop in a ticket and they WILL help you. The gang over there is really good about that!
For me 80% of the apps in Pinokio either fail to install or don't work at all. It also sucks that you wait 15 to 20 minutes for something that doesn't run at all. I got the same error as you have here before.
@@user-cz9bl6jp8b That's a shame. I was really hoping it would work as advertised. Whoever said using AI is easy just doesn't know what they're talking about. lol. I spend a good part of the day trying to get this to work, time I won't get back.
Pinokio caused trouble from the very start, and I would not recommend it to anyone who already has other AI installations like 'Stability Matrix' running. If they don't make it much more stable and reliable and isolated from other stuff OR respecting already present conda and git installations, I will never touch it again.
Make sure you are not on any windows 11 dev or public beta’s Pinokio has major issues on those every single update windows pushes. something about how pinokio works gets blocked by hard coded windows 11 security.
Looks amazing, great vid as always Tim… annnd an hour later I couldn’t get it to work though followed ever step. On my MacBook Pro Sequoia, but my side screen options diff at the point of ‘Ui Default’ (that was not there for me) … not surprised, me and local installs for top notch Ai content creation just never works out being compatible though I keep tryin
Huh. Racking my brain as to why that might be…you’ve got the latest version of Pinokio, I presume? Mayyyybe hit the “reset” at the bottom (which actually re-downloads it?) or try the update? Actually, try the update first, and then maybe Reset if that doesn’t work. Beyond that, I’d recommend checking out the Discord. They’ve got a pretty good support/help system going on there
Pinokio runs on my Pi5, wow. Been using Ollama to run LLM's, this looks like another to install and run AI stuff. Not sure how much works yet but but it seems simpler than other methods.
I think of the one click installers, Pinokio is kind of the king, they’ve got it the most together. Things just tend to work (although, obviously, sometimes you’ll get some fails!) Best bet: keep an eye on updating Pinokio itself. It doesn’t automatically update.
Another awesome video, Tim. I have been looking for something like this and I can't wait to try it. I've tried paid services as well as free tools like Forge and Fooocus but I can't seem to get them to do good face swaps. Hoping this will. And yes, don't be evil. Use responsibility.
I think you’ll be super pleased. Again, as I pointed out in the video, it still has limitations (don’t they all?) and make sure you check out the tips I popped in the description!
Hi Tim! i would like to politely request that you explain in more detail what the app or feature of the app is for. In the section about the face swapper section involving expressions (?) I wasnt able to understand the practical use for it as you zoomed through the explanation. Otherwise, great content! Thank you!
Ah, yes- sorry, I might have breezed though that part, since I covered it in an earlier video. If you go back to the video I did on OpenArt (the thumbnail has a girl with blue hair, I think?) there is a section that covers this feature. It’s not quite controlled the same, but it is pretty much the same feature. Everyone puts their spin on it, but the underlying technology is the same
Excellent!! Yeah, I really wanted to settle the issue on the whole Windows Defender issue. I love Pinokio and want more people using it-- but, I get it-- it seems malware-y at first if you aren't plugged into the whole scene. Hopefully we can use this video as a resource when people question it in the future. Or, once our Peanut friend decides to sign the dotted line!
Thanks Tim. Excellent video as always. Managed to download Pinokio (never knew it existed), as there was always 'stuff' i wanted to play with for fun on local but had not idea how to download/install - until now, thanks to you. Appreciate these tutorials. Regards from Ireland.
just installing now, there were some requirements for installation, are these going to be installed automatically or will i need to find all of these required programs first
Love it! Just rewatched it a few days ago with the kids-- since, given that Jenna Ortega is in the second one (and "the mom from Stranger Things"---sigh) they want to see it. Kids, man...just...sigh. So-- even if the new one sucks, at least they got to see the OG.
It'll basically run on anything-- it just might take longer with less horsepower. It's one of the things I love about Pinokio in general. Most apps will run on a Dell Potato. It'll just take a lot longer. And the machine might get stupid hot. But hey: Also makes a great place to warm up that Burrito!
@@TheoreticallyMedia so my idea is ,use minimax and create content with famouse actore faces, then just you this one and change to Ai genertaed faces. minimax create good video with famouse people faces, but not with AI generated from text . so sound to me good option
agreed on version 2. There was a general feeling of a "copy/paste" kind of vibe in that one. Version 3 seems MUCH better. You still get some wonk from time to time-- garbage in/out, as it were-- but overall, MUCH better.
i always find upscalers turn really glamorous ppl into regular looking people, its very frustrating, i may try the ageing slider to smooth out the heavy detailed textures it puts in the face.
Haha, I had a whole bit that got cut from this video, but I went into the channel's analytics and apparently in the last 28 days, there have been 18,458 views on the channel where someone was running Linux. (Yeah, RUclips tracks everything...) So, I had this whole slew of side jokes about Linux folks-- I think the next line there was "that's right, all 18,458 of you!" ...It got cut for time...and really, because no one was going to actually care! haha
No matter how good Facefusion gets its still hit and miss when detecting faces at certain angles and distances, whereas for some reason, using Reactor node in Comfy has a much better consistency.
They did add an interesting thing (that actually got cut from this video) where you can mark degrees (0, 90, 180) for the source-- so, that's helpful. I'm just still amazed that we're at a point where all of this is happening via a one-shot image. Man, the "old" days of having a folder of 300 images weren't that long ago!
I mean, I think it gets there-- Best can be subjective. And sure, I totally agree that there are crazy spaghetti Comfy workflows that might do a bit better-- but, that's also the juggle of doing YT videos, maintaining that balance of ease of use. To be totally honest, even Pinokio can be a big ask. Mark my words: I'll be getting tech support questions in this comment section for the next six months.
It can-- I didn't go into it in this video, but it basically does it via the UI Workflow at the bottom. To be honest, I didn't dig into that part too deeply, but you set it up so it runs twice: Once with one face, and then it circles back to do the other.
After you watch the video, I've got some extra tips that I missed in the video in the description! So, make sure to check those out for even better swaps!
For better face consistency make sure you've got the different angles selected. You can also enable 'many' on the face detectors although this will slow down the processing. Another issue you may have had is playing with the face distance slider.
Where can we test the Advanced Live Portrait ? I think I missed your video on this one
Do you have a video that touches on the other tools that come in the Pinokio install?
The darkest timeline gave me a good chuckle 😆
haha- same here!
Use pixel bost without face enhancing better quality and still think roop with faceset is better
Had some decent success with also running it in Topaz after with the Iris Model. That really seems to be a cherry on top!
deepfacelab the best but takes longer
tool is ayt, not perfect, which in this age, without perfection, it's useless, but good start... thank you.
Totally welcome! And I get it-- but here's a funny thought for you: Soooo much media consumption is on mobile these days-- and, this kind of stuck me the other day as I was watching one of my own videos: On mobile, a LOT of those imperfections kind of vanish.
I know most of us look at all of this with such a keen eye for detail-- but, the larger world (sadly, maybe?) really doesn't see it.
Yes - DO NOT BE EVIL! :)
Wish it didn't need to be said...but, yeah: The Real World.
Cute, but Deep Face Lab is still superior. :)
agreed
FaceFusion is such an annoying mix of greatness and noob programming. Its AI algorithm/processing: Greatness. Its code: Noob. It writes every single frame to disk as PNG and processes them step by step, rewriting the data over and over, wasting your hard disk's durability. It's an SSD wrecker. For a 1080p video this can easily mean 400 gigabytes of data written to the disk. It makes me not want to use it. The proper design would have been to read 1 frame at a time, process it in-memory, and send that frame to the video encoder live, without ever landing on the hard disk in-between.
sad to read that..
If you'd like to improve it they have a active discord, the 'early adopters' section is for FaceFusion Next with experimental features. FYI I have no idea how your using 400GB of space unless its a v long 1080p video with enhancements? Maybe look at some of the other more in depth tutorials if you've not done so already
@@jonogrimmer6013 There is no point to invite people like him.
fork it?
fork it?
According to AI, I need to adjust the age gap of older women that i would date. 😂
Cool! It works really well! I always recommend that people use a separate computer to run any of this AI stuff. It's the best way to avoid exfiltration of personal information and password stealing Trojans. Remote control and unauthorized processing (i.e. bitcoin) are still potential issues, but since it's on separate computer, it's something you can monitor and deal with in the future.
Time, and I say this with sincerity, you'd be a great comedian. Great video as always.
Ahhh, thank you! I mean, my whole take since the start of the AI Explosion is: This is so crazy, we have to laugh!
@@TheoreticallyMedia Reading my own comment again, I can't believe I called you "Time". Apologies 😅
When 33 yrs you reach, look so good you will not.
Too cool! Thanks for the (as always) excellent tutorial. I'm wondering if putting an old man's face on a Paul Rudd video would BSOD the computer(???)
Haha, that just might break all of Space/Time! hmmmm, might be worth it!
13:35 MOST IMPORTANT PART
Thanks for the video Tim. 👍I see a handful of gems in an ocean full of horrible misuse. Shame. How will I convince my wife I'm _not_ cheating on her! 😆
Sigh. I know-- it's true. It's such a double edged sword. The best I can do is encourage to not go Evil!
suppose there are multiple people in the video and we want to change various faces then it works or not? Similarly, if there is 2 or more people in the frame then how does lip sync works?
You should have targeting options (right person, left person) on the right had side.
At the bottom you’ll see a pull down that says “instant run” or something like that. You can use that to schedule “jobs” where FF3 will run once w/ one person, and then again with the other.
TBH, I didn’t have time to play around with jobs for this video, so try it out and let me know how it works!
Very interesting. Thanks a lot dear Tim. I try to install but the installation stops with an error when trying to install Microsoft Visual Studio. Anyone has an idea how to fix. Thanks in advance.
Best bet here is to hop over to their Discord. Lots of very smart folks over there who can help. If you put in a support ticket, they're really good about helping out too. It's funny, as a community driven project, they're actually better at support than most companies are!
It seems that you can't install it if your windows user name has a space in it. Mine is my first and last name and it fails to find my user folder because it is only looking for my first name.
You should be able to point it somewhere else. Check in the settings. I have it installed on an external drive. Granted, that external drive is named T7 (the Samsung portable SSD that everyone seems to have) and it installed fine.
I already have Roop Unleashed installed. Is face fusion better?
Roop is great too! In my opinion, I lean to FF, but do try it yourself to see what you think. It’s free!
(Plus, tons of other stuff on Pinokio if you aren’t already on it)
What is your pc specs ?
I'm on a Mac. I know, one of the few in the AI space that runs mac. I just have a (now older) M1 Macbook Pro. But, running fine.
To be honest: Pinokio and FF3 will run on about anything. Even a Dell Potato. It'll just take a lot longer.
How is this free? It looks 100% free with no small prints, so what gives? (Also... there is no standard size for croissants. As long as it is shaped as a crescent moon, it's all good).
Community Driven Open Source is fantastic!
And good note on the Croissants! I'll have to tell my bakery: MAKE IT AS BIG AS MY HEAD!!
I has a question. Is this like wav2lip where it bases on phonemes and mouthshapes, or is it english only?
Good question! I’ll try it out. I think I still have some Korean vocal stems from that Found Footage trailer I did awhile back
somehow i dont get the UI (DEFAULT) button after installing... _:(
Do you see Run (Default)? Try that--
Actually that was the problem, I think you missed that step in the tutorial 😅 anyway thanks !!!!!
the slider "face selector age" is not used to tell the AI the age of the person but it helps select people on the picture according to their age (like gender/race/distance selectors)
Ahhhhh- excellent! I'll pop that in the description!
why didn't I see any difference while reconstructing beetlejuice with your foto?
Haha, because I look as terrible as a minor demon in real life!
I tried to install pinokio on my M1 Macbook and it did not work out….. Anybody has it running on a M1?
I actually have it running on an M1. Did you run the patch?
If you’re having issues, best I can say is open a ticket on their discord support. They’re really good about helping out.
Again: this isn’t a company, it’s a community…which means the people actually care and want to help!
This is not the best face swapper, a combo of Reactor, Ipadapter and controlnet beats this by far.
Yeah, the take on this was more along the lines of "outside of Comfy" and more directed at the intermediate or starter level.
People apparently take their faceswappers pretty seriously! haha- I updated the title and thumbnail and removed "Best"--
Hey, at least I didn't say "Everything Just Changed!"
WOW, this has come a long way from what it used to be, it looks really good.
I *STAN* Pinokio (And Face Fusion) 😎... Great video! 👌🏽🤗
Thank you thank you! It's so good! And hopefully, reaching out to our Peanut friend, some more folks will join us-- at least, those who were on the fence w/ that WinDefender warning!
@@TheoreticallyMedia Yeah, it's good that you explained that 🤓. Hopefully raising awareness on the reasoning behind it will allow "Mr. Peanut" to stick to his original guns, too 😅
Great update! Have you discovered something that handles multiple faces in a video for swapping in different faces for each?
So, I didn't get into it for this video-- namely because I didn't have a lot of time to experiment with it, but at the bottom of FaceFusion, you'll see something that says "instant run" or something of the sort. Sorry, not at my machine...
Anyhow, that's a pulldown where you can use "Jobs"-- this will allow you to run FF on one character, then after it is done, it'll go back and do the other.
Again-- didn't have time to dig into this feature, but it is there!
Akool
why dont I get the CUDA option, only CPU is available
That depends on what graphics card you have. The CUDA option is there if you have an Nvidia card. Don’t worry about it though, those options just act as a speed boost by using the GPU. It’ll still work.
@@TheoreticallyMedia I do have a Nvidia gpu, rtx 3070 ti, but it doe not show up for me, do you know any fix?
On the pc: 1) Do you have both CPU & GPU selected? 2) There's also a 'tensort' option on windows. Should this be selected as well/on it's own?
I just turn them all on! Try it out, if it’s faster: do it!
@@TheoreticallyMedia Good call. I did that and it worked! Many thanks.
I'm getting an error while installing the "cloudflared" part of the facefusion installation, and I just can't get any further. Do you know anything about this? The error message goes by so quickly so I'm not able to read what the issue is...
AI Face Swap IO is free, done in browser, does not require an install. None of you should be installing random software you find on the Internet. Just use the Web stuff.
Some prefer the local option. Some like going web based. Its just good to have options.
@@TheoreticallyMedia In my opinion, inviting your followers to download possible spyware off Internet randos is going to screw you over eventually. "I spoke to the owners and they seemed like swell guys" is not the kind of antivirus you think it is.
@@dirtydevoteewhat criteria have you been using for all the software you have installed locally on your computer?
@@dirtydevotee Most software, whether open source or closed source, has bugs and exploits. Secure software is exceedingly rare these days. Even your hardware can have exploits and malware running under the level of the operating system, nearly impossible to remove.
@@ILoveTeles I'm on Linux. So if it's in the Linux Mint "software manager" on my taskbar, its allowed on my computer. It's 2024. There is literally nothing else that you can't find on the Web. Photopea, Dezgo, Krea, six video providers, and hundreds of everything you can get on Git. What are you people doing risking bricking your machines for a cheap thrill?
yep, dont use it for evil.
I'm enough of a realist to know that I need to say that at some point. Will it stop people? Probably not, but hey-- doing my part!
You cant use it for evil since its censored
I've been using AIVideoFaceSwap for some time and am really happy with it. The latest update which allows a quality setting of 1.5 (previously 1.0 was the max) also handles head movements and characters making faces very well.
Only disadvantage is the processing time, a 4 minute video took a good 4-6 hours on my 3080 at the highest quality setting.
you didnt tell us how to use a cgi face over a personal video. can this be done? in the promotional video it shows people with robot heads
It can be. It'll take anything (within reason) and swap. Like, people have asked about swapping animal faces...that....while it'll do it, looks horrifying. So-- within reason is what I'd say.
@@TheoreticallyMedia can you do a video on how to stuff like that? Because im using it and it doesnt seem to even register anything not totally human.
Nice tutorial. Despite using AI tools every day, this stuff keeps blowing me away. And it's kinda creepy too.
Should directml be clicked?
You're so funny! Btw, I have a question about a face leaving or entering the frame, and I wonder if you can advise me on how to avoid "glitches"? Or do you have a video that talks about it?
Installed it manually now, without Pinokio.
Works, but I always get totally blurry output, no matter how crisp the input images are. Can someone here address this? Quite sure that I am not the only one with this issue...
Sorry Pinokio is a No-Go - it doesn't notice or respect already installed conda and git versions and it even creates trouble if you want to uninstall and reinstall it after it stops to work. Which it did on my system for no apparent reason after a few month of not using it.
I'll say, that's one aspect of Pinokio that tends to be a bit wonky-- the whole uninstall/reinstall thing w/ previous version. I tend not to jump on any updates w/ any software-- but, w/ Pinokio, I tend to wait a few days or longer for bug quashing.
But, at the same time, it's an open source project, and not commercial or anything-- so a lot of leeway.
If you're having issues, high recommend you go let them know in the Discord. They can't fix it, if they don't know about it!
@@TheoreticallyMedia
I've got a massive bunch of AI tools installed manually (no problems being a computer scientist) and also use 'Stability Matrix' as a launcher of my main AI GUI like A1111, Fooocus etc. I guess it makes more sense if I stay on this path instead of risking Pinokio messing with my perfectly running stuff.
it install things in a enviroment. doesnt mess with your pc. never had a problem with it.. installed (and uninstall) MANY apps many times
I was hoping you'd show if it works in different media styles. On the extreme end of things, can it face swap /fuse from a real photo to a cartoon / anime? I remember a Corridor Crew video taking faces and putting into different art styles, like Yoji Shinkawa (Metal Gear Solid art). But that was about two years ago (video title: "Is this the death of VFX").
I remember that one! I’ll have to do some test, but I don’t think that this is where FF’s strength is. You’d likely end up with something that looks a little cursed. It’s more meant for realism.
But, I’ll give it a shot! Why not?!
@@TheoreticallyMedia I, and probably the rest of your viewers/subscribers, appreciate that you do all the AI platform testing, and respond to comments so efficiently. It's one of those things I would like to have more time to do but thankfully you're here. Thank you.
@@TheoreticallyMedia I tested this... Animated cartoon video of a real person made using Flux + KLING and the FF3 for swaping the cartoon face with real face photo of the person... Frankly it worked out amazingly well and looks gorgeous... The cartoon textures and lighting etc are very accurately preserved and the real face blends beautifully.. making the whole thing look like a very high end studio cartoon movie
how long does processing take, its at 1000,2 and its only a 32mb video with 964 frame
Make sure it's using your GPU properly. If it's running in software mode it will take forever.
That dev either lied to you or you didn't understand it and communicate it bad but getting rid of that warning has nothing to do with joining any "wall garden", He could pick any and I mean any (not related to Microsoft) legitimate code signing certificate from a trusted Certificate Authority (CA) like DigiCert, GlobalSign, Sectigo, etc., and that will make the software trusted by Windows and avoid the security warnings.
Awesome video! Can this be used for animation?
Great overview, I’ve been using insight face swapper, since you featured it and this a big leap forward for me, even the facial expression palette…I just have to get over the install, but thanks for the install simplification and general starting point settings. Much thanks Tim.
PS. AI audio has been sleeping or I’m late to the party , I hope we have this much control with real time audio mangling and transforming , although vocaflex has been doing a great job of sound imprinting to create new vocal voices, but I would like to see more sound synthesis that really can handle morphing sound in real time , kinda like realtime synthesis with a synth plant engine. 1nce again cheers Tim.
Are schools having problems with CSAM ?
Thanks Tim. Do you use comfy ui much after finding tools like Face Fusion 3, or do you find that Comfy is still the ai filmmaking tool with the most control?
Does anyone have a suggestion for a good AI platform to match green screened actors to background footage, lighting , colour grade etc? Viggle and wonder dynamics do a great job with CGI characters. This actually does it with faces. I would think matching footage of actors would be easier?
From now on I will rate AI on a 'Diddy do it?' score (higher is better).
This is a 4
So he definitely did it.
Rachel Weisz finally about to star in the Mummy 3 as she always should have.
Fun is one thing, evil is when you start making money with it
Oh, I think you can make money with it in certain cases: like, if you’ve got your own film you’re using it for…like, a remake of Face Off (haha, the first thing that came to mind), and you’re using your own actors.
But yeah, I get your point!
@@TheoreticallyMediathe way I understood it when I looked into FF, a few versions ago, was that the output is very strictly non commercial. Only the code is under the MIT license. Or something like that, I’m (clearly) not a lawyer. Which is a bit of a deal breaker if like you say you are an independent or hobbyist filmmaker.
One day maybe you get the chance to show your short film as part of a compilation on TV or somewhere but then you can’t because one of the tools you used has this limitation. This is also the problem with a lot of other tools, like ComfyUI nodes that you only use as a small part of your workflow. Yet if you are to follow the rules they dictate your entire outcome.
so if its 100 dollars that is fine, or what is wrong with you whoever made the headline align with thumbnail
❤
can run locally?
It can
sorry but rope pearl is better
I’ll check it out!
Sarah Tim Connor! HAHAHAHA! I think you did evil. Thanks for the demo. Good stuff, Maynard.
Haha-- "Come with me if you want to....oh, wait. That's what you look like. Uhhhh, maybe I'll just go back to the future..."
@@TheoreticallyMedia lol
Man, you're the best, thanks a lot for this kind of content! God bless you! 🙏🏼
A nice daily dose of Tim
kind of a light week this week-- but, I've got something PRETTY AWESOME brewing for Monday!
What are the requirements for running face fusion 3.0
it'll pretty much run on a Dell Potato. It just might run slowly. But-- yeah: You can use it on just about anything!
Where we are headed with A.I, in general, is nothing short of amazing. Can't wait to see where this goes in the next 10 years.
Keep up the great videos.
You're a good man for getting that Install messaged cleared up by the developer Tim! 😎
Figured that might have been the case when I installed it a while back, but good to get confirmation anyway :)
Haha, yeah- I get hit with that everytime I mention Pinokio in a video, so I decided to finally just get the word out-- so, at least next time I cover it and someone asks, I can just link to this video!
hell yes we love the pinokio gospel being pushed far and wide! another great video too
visions of chaos is better
hello sir. any free GOOD lip-sync NOT local?
I can't get Pinokio to run correctly. It keeps trying to install stuff and saying Error: ENOENT: no such file or directory, copyfile 'C:\pinokio\bin\miniconda\python.exe' -> 'C:\pinokio\bin\miniconda\python3.exe' I have both miniconda and python running so I don't understand what's happening.
Hmmm-- might need to fully uninstall and then reinstall. That sort of thing has happened to me in the past. To be honest: Best bet is to check in on the Discord. Pop in a ticket and they WILL help you. The gang over there is really good about that!
For me 80% of the apps in Pinokio either fail to install or don't work at all. It also sucks that you wait 15 to 20 minutes for something that doesn't run at all. I got the same error as you have here before.
@@user-cz9bl6jp8b That's a shame. I was really hoping it would work as advertised. Whoever said using AI is easy just doesn't know what they're talking about. lol. I spend a good part of the day trying to get this to work, time I won't get back.
Pinokio caused trouble from the very start, and I would not recommend it to anyone who already has other AI installations like 'Stability Matrix' running. If they don't make it much more stable and reliable and isolated from other stuff OR respecting already present conda and git installations, I will never touch it again.
Make sure you are not on any windows 11 dev or public beta’s
Pinokio has major issues on those every single update windows pushes.
something about how pinokio works gets blocked by hard coded windows 11 security.
Looks amazing, great vid as always Tim… annnd an hour later I couldn’t get it to work though followed ever step. On my MacBook Pro Sequoia, but my side screen options diff at the point of ‘Ui Default’ (that was not there for me) … not surprised, me and local installs for top notch Ai content creation just never works out being compatible though I keep tryin
Huh. Racking my brain as to why that might be…you’ve got the latest version of Pinokio, I presume? Mayyyybe hit the “reset” at the bottom (which actually re-downloads it?) or try the update?
Actually, try the update first, and then maybe Reset if that doesn’t work.
Beyond that, I’d recommend checking out the Discord. They’ve got a pretty good support/help system going on there
Can you use 2 reference images for better accuracy if you have a profile photo, too?
Wait..i had an option for default +job.....wuts that?
Does it not work offline without internet?
🫴 Face Off 🖐 ...
Greatest movie ever? Haha- it was subtle, but throwing in Nic Cage as Bond at the start of this video was a little in-joke.
I think you should arrange a Facebook/Whatsapp or Discord group where we the enthusiasts can post our issues to solve.
I know-- I need to do that. I had a Discord awhile back, but w/o mods it turned into a den of spam bots.
Looking for an alternate solution...
@@TheoreticallyMedia Glad to get your opinion. We will eagerly wait for the alternatives. 👁❤ Gracias
Is there a remote version of facefusion3? My pc is too old.
Pinokio runs on my Pi5, wow. Been using Ollama to run LLM's, this looks like another to install and run AI stuff. Not sure how much works yet but but it seems simpler than other methods.
I think of the one click installers, Pinokio is kind of the king, they’ve got it the most together. Things just tend to work (although, obviously, sometimes you’ll get some fails!)
Best bet: keep an eye on updating Pinokio itself. It doesn’t automatically update.
Another awesome video, Tim. I have been looking for something like this and I can't wait to try it. I've tried paid services as well as free tools like Forge and Fooocus but I can't seem to get them to do good face swaps. Hoping this will. And yes, don't be evil. Use responsibility.
I think you’ll be super pleased. Again, as I pointed out in the video, it still has limitations (don’t they all?) and make sure you check out the tips I popped in the description!
Can one use Facefusion offline?
I believe so. Funny enough, I’ve actually never tried
Yes, I often use it offline.
Hi Tim! i would like to politely request that you explain in more detail what the app or feature of the app is for. In the section about the face swapper section involving expressions (?) I wasnt able to understand the practical use for it as you zoomed through the explanation. Otherwise, great content! Thank you!
Ah, yes- sorry, I might have breezed though that part, since I covered it in an earlier video. If you go back to the video I did on OpenArt (the thumbnail has a girl with blue hair, I think?) there is a section that covers this feature.
It’s not quite controlled the same, but it is pretty much the same feature. Everyone puts their spin on it, but the underlying technology is the same
can't get past the terminal for the install. it just stops.
@13:00 you only have CPU selected.
Thankyou, I've been waiting for this. I love Pinokio!!!
Excellent!! Yeah, I really wanted to settle the issue on the whole Windows Defender issue. I love Pinokio and want more people using it-- but, I get it-- it seems malware-y at first if you aren't plugged into the whole scene. Hopefully we can use this video as a resource when people question it in the future.
Or, once our Peanut friend decides to sign the dotted line!
visions of chaos is better
Thanks Tim. Excellent video as always. Managed to download Pinokio (never knew it existed), as there was always 'stuff' i wanted to play with for fun on local but had not idea how to download/install - until now, thanks to you. Appreciate these tutorials. Regards from Ireland.
This is the WIN for the video!! So happy to hear this!!
just installing now, there were some requirements for installation, are these going to be installed automatically or will i need to find all of these required programs first
That should all install with Pinokio, that’s one nice thing, you don’t have to hunt down all the dependencies
When I install FaceFusion - it wont install cloudflared. Anyone know how to fix this? I am using MacBook Pro with Sonomona 14.6.1
That’s odd. I’m 14.5. Hmmm, might want to check with the Pinokio discord on this one.
Exciting!! Can't wait to try it!💯
Can this be done online?
It can be-- there are a number of services that do it-- but I think the angle here is this one can be done on a local machine w/o any subscriptions.
@@TheoreticallyMedia Yes, of course, it's just that I don't have a powerful enough PC. Thanks for your reply!
visions of choas is better than pinokio
The Juice is Loose
Love it! Just rewatched it a few days ago with the kids-- since, given that Jenna Ortega is in the second one (and "the mom from Stranger Things"---sigh) they want to see it.
Kids, man...just...sigh.
So-- even if the new one sucks, at least they got to see the OG.
another greate video. what GPU do I have to get for this? 12 GB? will do or need more like 24GB?
It'll basically run on anything-- it just might take longer with less horsepower. It's one of the things I love about Pinokio in general. Most apps will run on a Dell Potato. It'll just take a lot longer.
And the machine might get stupid hot. But hey: Also makes a great place to warm up that Burrito!
@@TheoreticallyMedia so my idea is ,use minimax and create content with famouse actore faces, then just you this one and change to Ai genertaed faces. minimax create good video with famouse people faces, but not with AI generated from text . so sound to me good option
@@Comic_Book_Creator you got nice video? Like I used tim gpt but still I get illustration type of videos from minimax .
@@usmansiraj-g8n no, I us custo gpt to creat the prompts
version 2 had very very low resolution images , and was therefore unusable , what your opinion with version 3 ?
agreed on version 2. There was a general feeling of a "copy/paste" kind of vibe in that one. Version 3 seems MUCH better. You still get some wonk from time to time-- garbage in/out, as it were-- but overall, MUCH better.
i always find upscalers turn really glamorous ppl into regular looking people, its very frustrating, i may try the ageing slider to smooth out the heavy detailed textures it puts in the face.
Ahhhh- that's a good use case for aging! I like it!
coreml isn't showing on mine
Hard to believe in: "Linux people can download it with no problem"
Haha, I had a whole bit that got cut from this video, but I went into the channel's analytics and apparently in the last 28 days, there have been 18,458 views on the channel where someone was running Linux. (Yeah, RUclips tracks everything...)
So, I had this whole slew of side jokes about Linux folks-- I think the next line there was "that's right, all 18,458 of you!"
...It got cut for time...and really, because no one was going to actually care! haha
No matter how good Facefusion gets its still hit and miss when detecting faces at certain angles and distances, whereas for some reason, using Reactor node in Comfy has a much better consistency.
They did add an interesting thing (that actually got cut from this video) where you can mark degrees (0, 90, 180) for the source-- so, that's helpful. I'm just still amazed that we're at a point where all of this is happening via a one-shot image. Man, the "old" days of having a folder of 300 images weren't that long ago!
@@TheoreticallyMedia It truly is going ahead by leaps and bounds, and fast!
Thankyou, this will save me money
Excellent to hear! Although-- haha, I know that money probably now get spent elsewhere! But still, hopefully it's somewhere fun!
WHY OH WHY waste time with this
in a few days or weeks some Ai is going to be on a platform that can do this easily
So don't waste your time people
your logic then why not wait till next year it will be better then right
Thanks Tim!
1000%!!
Thanks Tim!
1000%!! Thank you!
Maybe easiest to use, but not the best quality
I mean, I think it gets there-- Best can be subjective. And sure, I totally agree that there are crazy spaghetti Comfy workflows that might do a bit better-- but, that's also the juggle of doing YT videos, maintaining that balance of ease of use.
To be totally honest, even Pinokio can be a big ask. Mark my words: I'll be getting tech support questions in this comment section for the next six months.
looks a lot better than what I was getting with IPAdapter
there are way better configs when using facefusion! much better results than showed. just fyi
Drop tips! I'm posting them in the Description as they come in!
Well I hate to be the one to report this, but the tool is censored.
How is it censored? I mean, I don’t run NSFW stuff personally, but I’d be surprised if it did, considering it runs locally.
@@TheoreticallyMedia put any image in that has any nudity and its immediately blocks it. Its also in the github which says the model is censored
You can't just say that , at least explain what you mean by censored please ˆˆ
No to pinocio.
Can it swap multiple faces in one frame like Akool? Asking for a friend and his companions…ruclips.net/video/g0q72r-RXMQ/видео.htmlsi=9eKTxS-uzBxcyMMb
It can-- I didn't go into it in this video, but it basically does it via the UI Workflow at the bottom. To be honest, I didn't dig into that part too deeply, but you set it up so it runs twice: Once with one face, and then it circles back to do the other.
@@TheoreticallyMedia awesome!