The Craziest Faceswap I've Seen Yet / Midjourney's Future & Two New AI Video Platforms!

so face swapping and AI avatars have taken a pretty big jump I've got one to show you today that is really going to blow you away plus we're taking a look into the future of mid Journey at least a 12-month road map and it is headed in a pretty surprising Direction all that plus two new AI video generators that you're going to want to check out okay lots to cover let's dive in first off welcome to Studio B uh the main studio is having some ceiling work done so we're going to be here for a little bit but that's not important what is important is this crazy face [Music] swap this comes to us via AI Katana although I don't have a source on the original footage and obviously he is speaking in either Mandarin or Cantonese I'm not sure which uh although we do have a little bit of a translation on that coming up in just a minute while we have seen technology like this before I'm thinking like Snapchat filters uh this is kind of taking things to a whole other level uh the way that it looks convincingly tracking uh as he's eating uh is very impressive and then when he's sort of tugging on the cheeks as well uh that looks remarkably realistic now I do think that the capture footage up here is actually on a different camera than what's actually being generated uh if you note here um as he's kind of looking in this direction it doesn't actually match the angle of the uh deep fake version now in terms of what he's using and what he's saying id8 films the Channel's new unofficial translator uh said that he's pointing out the differences and advantages of what he's using versus current face whopping Tech and that this is a model that he trained there has been a lot of speculation that this is probably not running in real time rather this is a video capture run through face swapping software real-time face swapping does still have some issues even as good as it is face swapping here to Elon Musk uh we can see that well again it does look very good there is still a lot of like incoherencies and morphing kind of going on uh things that would not hold up to the smell test though this definitely looks like it has taken a significant jump up from the last generation of face swappers namely face fusion and rof I'll keep an eye on this one and I'll let you know what the process is as soon as we find out meanwhile if I go to Amazon and all the wigs are sold out I'm going to be a little disappointed in you all next up the next generation of AI avatars have arrived and this time they have emotions I am very happy I am so upset I am frustrated this one comes to us from Synthesia this is their new Express one model let's check it out now focus on my lips they're snappier and align more precisely with each word I say to you and my voice do you hear the difference so this is not a haen type situation where you will be recording yourself in order to generate your avatar but rather uh there will be pre-trained avatars for you to use which is understandable given that this is just a look at part of their capture rig I'll point out that I do have room for this in Studio B so if you're looking for an additional training location I'm here for you Synthesia so let's take a look at Express one to see how much more emotive these AI avatars have gotten I'm not convinced I definitely need more details I need proof of concept thank you thank you so much for hanging in there with us we really appreciate it we'll learn more about Synthesia coming up soon they have a launch event that is linked down below if you want to learn more moving on to some mid-journey news and essentially looking at their next 12 months in this week's office hours Nick sa Pierre reported that CEO David Holt said for the next 12 months it's about video 3D real time and bringing them all together to non-interactive World simulator then we'll add the interaction layer to it what does that actually mean well it's been speculated at least on the 3D side that instead of generating images in mid Journey you will instead be generating scenes at which point you will have full 360° rotational camera placement control over that scene a while back once again Nick s Pierre put together a up of what this might look like uh once again this is not actually mid Journey generated this is rather a Nerf that he ComEd into sort of an example uh but we can get an idea of essentially you know what this 3D feature in mid Journey might look like mid Journey step into 3D is really fascinating and made even more exciting by the fact that Alex Evans one of the co-founders of medium molecule has joined mid Journey as a principal research engineer media molecule was the developer of dreams for the PlayStation which was essentially a 3D creation engine one thing mentioned in the office hours was that 3D has been held back by lack of data but data collection efforts are ramping up which kind of reminded me of the orb kind of a long running mid journey I hesitate to call it a joke it's kind of more like an urban legend David Holtz has described the orb as a device which could generate and manage thousands of 3D rooms and it does seem like mid journey is very serious about the orb as they have hired Ahad aboss as their head of Hardware uh Ahmad was one of the driving forces behind the Apple fion Pro and we definitely know that 3D in mid Journey will come before video although I do have two new video tools for you to check out coming up in just a minute as a quick mid-journey side note I recently did a sort of beginners mid-journey course for sem Rush this video is not sponsored by them it was like a whole other gig that I did but it is a free course on getting started with mid Journey if you're interested in checking that out the course is not just me there are a number of other instructors in there but I'm the only one that handles the mid Journey portion of it of Studio a look at you so if you're interested in a free video on how to get started with mid Journey that link is down below routing out with the mid Journey news they recently released a new feature called style random as the command name implies this randomizes your style and I think it was met with a lot of like eh what are we going to do with that but as it turns out when you start playing with it not only is a lot of fun but it is pretty useful for an example we're going to take an image that I previously generated this is a gum shoe detective rough but handsome I don't know about the handsome part but definitely got the rough in there uh sitting in his cluttered office holding a cup of coffee case folders all over his desk light streaming in through closed blinds cinematic uh still film grain so I am going to remove cinematic still and film grain because we're using style random and I don't want to call that out and all we need to do here is isy the Comm in dash dash SRE space random and that gives us an image that is completely different stylistically and objectively much more hilarious everything about this screams like lost Monty Python skit directed by Wes Anderson uh but I can say I do really want this painting for Studio B writing it a few more times we can end up with looks ranging from anime to a very Stark graphic novel look uh this one in particular I really like uh very much reminds me of Mike mingola to a style like this where I don't even know necessarily what I would prompt to end up with this so obviously the real fun in Sr random is just kind of rolling the dice and seeing what you end up with but where it actually becomes very useful is when you do stumble across something that you like you can then continue to use that style for example taking our super dark red splattery image here um all I have to do is hit this srf button and I can put it next to a prompt like an old man standing by a tree in a city park uh and now it will reference this style for that image we end up with an image like this which kind of looks like Santa on the run from a murder he just committed moving on we have two new AI video generators uh the first is morph Studios which we have previously taken a look at uh this is in beta but the beta is just beginning to roll out now morph style definitely leans towards kind of an animated look although there is some pretty interesting things that you can do that we'll take a look at in just a second uh you can also upload character images for consistent characters and styles it also does have a lip sync and sound feature as well but I think the UI is the interesting aspect of this Heather Cooper got Early Access to morph and kind of has a screen share of what the UI looks like and this was the thing that I found fascinating in that initial look is that if you look here there's kind of like this node-based structure to it so you can prompt reroll for different styles and then choose aspects of that to connect to the next shot and or node as you can see here we also have the motion brush tool and then ultimately we can connect different nodes for different exports so it's a really interesting workflow I'm very curious to try it out Heather also gave us a interesting look at two different styles in a split screen obviously it's split top and bottom but uh what's fascinating about this is that this shot is 10 seconds long I don't have access to morph Studio yet but the beta is now and they are beginning to roll access out so as soon as I get it I will have a deeper look at it for you rounding out we have Nim video this is another one that is currently in beta and has a pretty interesting workspace as well just walking through this very quickly it does appear that we have options for style and character uh consistent characters just seem to be a thing now uh camera motion motion strength obviously uh sound and lip sync as well some other features called out by Nim video include image to video uh video restyling an upscaler and the ability to work in layers as well as motion control uh Regional editing which I presume would be in and out painting and lip syncing as well Nvidia will also be using open source models for its platform so if you're interested in checking it out you can hit the link below and sign up for the beta so that's it for today I am going to go find something to hang behind me cuz Studio B feels a little bit on the empty side I thank you so much for watching my name is him

Channel: Theoretically Media

Views: 66,588

Keywords: Faceswap, deepfake, midjourney, ai video

Id: lARo9uc88zQ

Length: 10min 38sec (638 seconds)

Published: Thu Apr 25 2024