 Hey guys welcome back to my YouTube channel this is Daniel Rosal here bringing you today's video from my home office in Jerusalem. Very very pleased to wake up today to see I've crossed over the 800 subscriber mark here on YouTube 200 to go until 1000. Never expected to even get this far to be honest although I've been putting out these videos for a little while. Thank you for anyone who has subscribed to the channel to date and stuck with it I know the subject matter is super varied I mentioned before that I tend to that I intend in the future to break down into more niche channels or more thematic channels but right now I'm just having great fun learning a lot about video production meeting interesting new people interviewing interesting new people and bringing my camera out with me to interesting places. So speaking about interesting things and interesting people I got to mention on Twitter today from a very good friend of mine a guy called Peter Duffy from back in Ireland in the motherland and Duffy is a very very astute follower of technology he puts out a terrific monthly newsletter called How Curious and it's kind of an interesting synopsis usually of tech Duffy has seen a podcast he's been listening to he's an avid podcast listener and he's also a big non-fiction reader so he's very much a guy sort of aligned with my interest but he's a he knows that I'm really interested in artificial intelligence and this whole world of deep fakes because I may or may not once have published a book under a fake name and that was really my kind of eye-opening experience for me into the worlds of deep if you know fake faces fake videos and how quickly artificial intelligence is advancing I've also written a couple of articles I'll put a link maybe to one of the ones I've written on medium my predictions for that this market for AI generated fakes and deep fakes of all kind is going it's evolving very very quickly right now as you're going to see in a minute the tech is not wonderful because we have two things required for deep fakes ready one of them is credible synthesized voices or TTS text to speech that's been a tech sector that's been evolving in recent years it's getting very good but it's not quite there yet it always sounds a bit robotic and then the more recent development or the more recent tech we're seeing is websites like the one I'm going to show you that Duffy told me about it's called movio.la and again you can see the mouth the mouth movements look a little bit artificial but it's pretty good and it's kind of freaky so I thought in this video I would show you guys how this webs give you a little inside look now there's another website that I've been following this is movio.la and they've got a few templates but you can also create something completely from scratch another website that I've just been on my radar for a long a long time now is called this person does not exist there's actually a website called this x does not exist and the link to various projects is something using an algorithm I believe it's called GAN from NVIDIA and you can read about how the algorithm works here or you can just get fake faces now I'm going to hit the refresh image button and we're going to so what I think is really fascinating about this this whole world of deep fakes it's not any of these one tools by themselves it's the fact that when you put these ingredients together we're going to get a fake face that looks like a real person then we're going to take that fake person and make it into a deep fake video and we're going to create a video of a person that never existed right so it's when you start combining stuff that it gets pretty wild I'm going to just wait until I can get a little bit better they say the best type of face for this project for a movie or dot l a is a face that isn't smiling almost like a passport photo so I'm going to just keep hitting refresh until I get a pretty neutral adult face and then I'm going to make a video from it with a backdrop of let's say the western wall in Jerusalem because it's where I live you know what this guy is pretty decent I'm going to take him and then let's see what kind of a video we can make from this okay so I've gone ahead now and I'm downloaded my fake face from the internet I'm going to call this guy Dave now something that would be very cool if they added this feature was I'm just going to use this as is but if you have someone with a background you can just do this kind of manual cropping I'm doing now in GIMP and then put them on a transparent background and that will allow you to get an even more realistic deep fake video for now I'm just going to play with Dave's Dave's fake itself and see how this works okay so I've gone ahead now and done a pretty crude job at cutting out my fake Dave from the AI generated background so I'm going to go ahead and save this as a transparency you guys can see how well it handles transparency at least it's going to give the avatar a background to stand against instead of the background that comes with the generator and let's go ahead and plug that in so I'm loading up a Muvio here and you get options to create landscape or portrait based videos and I've gone for landscapes so you can also upload your own backdrops I'm going to go ahead now and put in a custom backdrop so I just took a random scenic backdrop I did google images and I went for one that was licensed appropriately under the creative commons so that's going to be my backdrop and you can also add different elements here and textual elements into the video for instance if I want to add overlays for whatsapp with certain frames of the video or background music I can also do that but let's just keep it simple for this example that's the background I'm going to be using now I'm going to go ahead and add in my fake avatar okay so I'm uploading Dave into my avatar library now as you can see Dave is just uploaded there as well as the ones that I can upload of myself or of my friends there are different avatars now they do give you a bit of guidance in terms of how you should do this and I'm going to call that up in a second but for now let's just take Dave now as you can see in terms of the transparency let me just just jump back to gim so you can see what I did you can see I did a bit of a scrappy job here this is actually kind of a bit of his hair against the background and other I guess clop crops of his hair but it still got this kind of white background showing there which to me kind of detracts from how credible it looks and it's obviously not a head head and shoulders one it one either but anyway that's kind of decent enough for the purpose of demonstration and what you can go ahead and do here is create a script and then you can pick between your fake AI generated voices right so I'm going to write a little script here I'm going to say hi this is Dave I'm not a real person I never oops I never was a real person instead I'm an I'm an artificial intelligence generated face that has been super imposed on a real background the purpose of this video is to see whether my face and background looks credible enough for your average viewer to believe that I am in fact real and something I like about the script writer on this is that you can choose to manually add in five second pauses and you can also change the speed there is a 60 second speed limit on how you know on how long it can run for and you can choose between your different voices so clearly this Dave is a man so I'm going to go for a fairly neutral voice it's not the most amazing AI voice I've ever heard but for the purpose of this test it should be decent and I really like this a deep fake creator I guess that's not really what it's called this AI video based creator because it's really an intuitive user interface as well as choosing your different elements you can choose to put on subtitles so I can choose to have my subtitles here I'm going to move my fake tongue head a little bit out of the way and now I'm going to just go ahead and click on submit and this is then going to go into the rendering queue now in order to get the watermark out you do need to pay for the service for this demo I'm just going to show you what it looks like with some watermark inside of it and as you can see it's now rendering so this is a kind of graphical rendering process that happens in the cloud presumably it's quite compute intensive as you can see we're only at five percent after you know 10 or 20 seconds so it's going to take a few minutes to develop your your deep fake avatar but we're going to see the output in a second one last thing I wanted to say really about what's scary about this technology to me so as my friend Peter Duffy remarked the credibility of the deep fakes that are being created now with this they're not amazing I don't think anyone's really going to be tricked but we can expect this whole market to advance very quickly now there's not really such a concern about you know deep fakes your friends being able to deep fake you maybe there is in the sense of pranks I think what's very scary about this technology is a potential for hackers or you know social engineers to create deep fakes of world leaders or presidents or celebrities and create fake accounts and put those two elements together so again I'm not I'm not encouraging by any means anyone to use this tech for malicious purposes the white hat approach would be well if you want a business and you and if you have a business and you want to make a simple what's called talking head video or you just have a person talking to camera kind of like what I'm doing instead of hiring a real actor and giving them a script you could use one of these avatars so Dave is rendering here and while that's going on I'm just going to actually create a second video and something I like about this is that you can render simultaneously so let's take a video of this guy he's pretty you know corporate looking so this might be I guess good for one of those corporate videos and I'm going to go ahead and drop in a background from Pexels of an office setting and we'll see if we can create a perhaps more credible looking fake talking head video than we did in our in our last one okay so just in order to try challenge movio la's technology was something that wasn't in their default stock library I went on to google images in the end and I searched for Tel Aviv so this is going to be a little background at Tel Aviv now again something I really like about this system is that it's really intuitive so if you've used any kind of image editor whether it's GIMP or photoshop or something like that you're probably very familiar with working with layers so what I'm going to do now is bring my fake speaking person or let's call him Dave too up to the front now as you can see I've chosen it's probably not the best background because he's in a black suit and it's all it's quite a dark background so the contrast isn't amazing I'm going to pop him down in the bottom right corner so he doesn't look totally out of place this almost looks like a kind of one of those virtual green screen screenshots except of course that our avatar isn't real so I'm going to give him another script to say I'm going to say hi this is Dave too and this is the second proof of concept for what are known as deep fake videos that Daniel has created for his YouTube channel and again the pauses the kind of limitations of the avatar speaking are going to be that you know sometimes if you use foreign words like Tel Aviv is a word in Hebrew right so let's see how well it handles Aviv I'm here today in Tel Aviv and it's not as hot as it was during the day thankfully you should totally subscribe to Daniel's YouTube channel which has just passed 800 subscribers he loves sharing cool new technology with his friends and those interested now something that's also pretty cool about the voice you can use to go along with your avatar is that as you can see here on the left panel there are options for different languages so I click into Hebrew for instance we have Avril now I know the audio is coming through my speakers not through the microphone so I hope that was audible but it's text to speech Hebrew and I have to say it's decent I wouldn't say it's either better or worse than the English versions right there's one option in Hebrew versus many in English it does sound a bit robotic but it doesn't definitely sounds quite understandable so I'm going to just go for a pretty neutral male voice we're going to go with Matthew which is one of the US voices I've written I have a hundred and fifty eight eighty four characters there's a lot of room to work with here and again you get a little timeline for those who are used to video editing I can add little features you can add elements and you can add text at certain points but just to keep it simple let's go for a talking talking head video Dave's going to give us a little speech in front of virtual Tel Aviv and I'm going to add that now into my rendering queue so the good news is that while my second video is rendering my first one's already ready so I can download here with a watermark for free something quite cool is it also gives you a URL to share so I can click on this share link share this on my Facebook or you can download the mp4 and of course the good thing about downloading the video is that you can then edit it in a video editing program what I did using Kaden live which is a video editor for Linux is I reduced the speed to 80% with pitch compensation and I think it came across a little better I would say that these tend to be a bit fast so my first avatar video is ready my second one is in queue I'm going to go ahead now and pause the video and then let you guys see how the output looked like I'd be very interested in hearing your thoughts on this avatar or deep fake technology coming from movio.la if you've seen better technology for this kind of thing that you like please let me know in the comments if you have thoughts about what the advent and rapid advancement and sophistication of this technology is going to mean for both white hat marketing uses and black hat uses for disinformation misinformation and all other kinds of mischief that can be achieved when you can take fake people or real people and make fake videos out of them and share them with the world as if it was something really done by that person if you have thoughts on what all this is going to mean for the future of the internet for the future of society and the sharing of information please let me know in the comments too because this is something I've really been thinking about and I must admit I'm a little worried about it because this technology is definitely rapidly coming to a point of good maturity and I expect in the next couple of years these synthesized voices will be a lot better these the fake speaking technology will be a lot better so what's kind of proof of concept and leading edge today is going to become relatively mature tech I imagine a lot sooner than any of us imagine thank you guys very much for watching this video I'll leave you with my videos of the deep fakes so stick around until the end if you want to get more videos from me please subscribe to this youtube channel thank you very much for watching hi this is Dave I'm not a real person I never was a real person instead I'm an artificial intelligence generated face that has been superimposed on a real background the purpose of this video is to see whether my face and background looks credible enough for your average viewer to believe that I am in fact hi this is Dave too and this is the second proof of concept for what are known as deep fake videos that Daniel has created for his youtube channel I'm here today in Tel Aviv and it's not as hot as it was during the day thankfully you should totally subscribe to Daniel's youtube channel which has just passed 800 subscribers he loves sharing cool new technology with his friends and those interested