The Craziest Faceswap I've Seen Yet / Midjourney's Future & Two New AI Video Platforms!
Вставка
- Опубліковано 22 тра 2024
- Pretty wild week in AI! I've got a look at the wildest Faceswap/Deepfake I've seen yet. This is one you need to see! Plus, a new AI Avatar platform that features AI Characters who actually emote!
From there, we'll head over to Midjourney to check out their plans over the next year, which include generating full 3D Worlds and Video. We'll also check in on their latest feature: Style Random, which turns out to not only be a lot of fun, but actually useful as well!
Plus, we've got a look at 2 new upcoming AI Video Generators!
LINKS:
Mystery FaceSwap: x.com/ai_katana/status/178297...
Expressive AI Avatars: www.synthesia.io/
Morph Studios: app.morphstudio.com/
Nim Video: nim.video/
FREE Midjourney Lesson by ME!
www.semrush.com/academy/cours...
00:00 - Intro
0:31 - The Wildest Faceswap I've Seen
2:38 - AI Avatars with Emotions
3:51 - Midjourney's Next Year
5:05 - The Midjourney Orb
6:15 - Midjourney's new Style Random Feature
8:08 - Morph Studio
9:36 - Nim Video - Наука та технологія
Thanks, Tim. Brilliantly helpful updates as always :-)
Thank you so much!! Much appreciated from Studio B!
Nim and Morph looks cool! I hope that I’ll have the ability to check it out soon! I am also so excited about the Midjourney news, it would be awesome to do everything in one place, it would really streamline the process 😊
I’m super fascinated by what MJ is up to! I know a lot of folks like to proclaim them “dead” but I always think of them as a slumbering giant.
David is a smart dude…a bit weird? Yup, and I mean that as a compliment, but also with a pretty grand vision of what MJ can be. All of which is to say, I’ve got some chips in his corner for sure.
Always the best Tim! Said with love,studio B needs a paint job :)
But anyhow,, its YOU we come to see and hear Thanks as always!
Haha. studio B is going to be a work in progress. I’m sure once I get it dialed in, it’ll be time to move back downstairs!
Also, Studio B is the kids project hangout room! They want me out too!
I love the style transfer feature showed at 8:00, there's so many different styles you get with AI images and it's so cool that we can now apply them with our own ideas!
10:35 - I think it was last year but I recall there being either an interview or something about the potential future of MJ and one of the comment was basically a "what if you could generate an entire 3D world using MJ" and my brain immodestly thought of something like the Nerve Gear in Sword Art Online but you are creator going into a blank slate and using prompts through programs like MJ to generate and build the virtual environment. I don't see it getting that crazy in the near future but I am here to see what MJ does with 3D Generation and where it could go.
10 seconds on Morph is very appealing! And can I say I love you even more for "whole other"ing instad of "whole nuther"ing.
Lol you're 3rd youtuber I'm subscribing who is going through "modernization" of "studio A" 😆 is there any particular building where all of you are hiding? 😂
Thanks for the video even in these difficult times. Voice wasn't issue for me.
Haha, spring renovations in full gear! In my case, Studio A is basically a giant barn. It’s an awesome space, but the roof is vaulted and uninsulated, so when the sun beats down in the summer the AC can’t keep up.
So, a few weeks of futzing in Studio B will payoff in the middle of the summer for sure!
Thanks Tim, great sum up as always!
Thank you from Studio B! Haha. Gonna spend today obsessively futzing with a different camera angle!
Fascinating round-up of the latest developments in AI.
appreciate it! Thanks so much!
Great video, looks like some interesting stuff is coming out.
Tim, Good sound is studio B, very nice!
Awesome updates thank you 🙏
This has been done in an even greater detail, for over four years now on a phone app, it just takes a little bit of tricky, but it can be done totally in real time and even more convincing than this. I think there’s something else going on here I don’t know how they did it on the phone but it been tested, on even an old iPhone anyway good luck with this. It looks better in terms of ability to modify the code Xa
I’ve seen some really good phone ones as well! I mean, even some of the Tik Tok filters are super impressive. For the older models, I’ll say, one that that helps in their favor is the lower resolution of 720p face cameras. A little smudging around the pixels really helps sell the illusion.
Yup and they could move the borders around a bit. They could have something very powerful. It has the age the age thing on point I still don’t know how they managed to do it. Everything on the phone even off-line great topic.
I usually focus on your excellent content and humour. So, although I'm thankful you shared this, as always, here's what really caught my eye:
0:55 'this technology have zero positive outcomes' (comment)
Although I don't agree with that entirely, it is _very_ disturbing what this will _undoubtedly_ be used for, and much too often.
Oh agreed! And good catch there! I actually (subtile) left that comment there instead of cropping it out for that very reason!
Wow that's crazy the face swap . We won't know who we are really talking to online anymore.
Hmmmm i'm sorry but as someone who works in the vfx industry I don't think the faceswap is legit, there are too many odd things about the faces. He has an invisible eyeglasses shape throughout on his face for one thing which seems to indicate something has been edited on the original plate which there would be no reason for. Also, some portions of the frame are flipped from the original but other portions are not flipped. Alot of things about this say this is a faked deepfake.... enjoy your videos by the way thanks for finding this stuff!
Thanks for your analysis! I’ve been hearing a lot of speculation stating the same. So, either this is a model that is really (really) impressive, or he choreographed the whole thing- which, I’ve got to say, is practically a magic trick all its own!
Tim, in your "--sref random" example where you then use the sref seed button to continue with that style...
Q:: What if I like that dark red splattery style, but I want to apply it in another color to create a variety of different colored images that share that style?
Thinking out loud here, what I would try is to take the red splatter image, run a Remix on it, and feed in keywords to the prompt like “blue hue, (dash,dash)No red”
Might take some futzing, but once you have something, use that as the new style reference.
@Theoretically Media would you care to make a video or tutorial about lips sync and face expression for the anime or animated charater we generate from ai ? Like where we upload the picture and audio and it got generated as lips sync I tried few in Google colb but they are not soo great they make lips blurry, please help
... signed up for mid-journey 12 months anticipating next moves in their tech ... should be interesting!
Excellent to hear! They’ve been one of those subs that I’ve kept since the start of this whole thing. Doing the channel, I obviously have to sign up for a bunch of different services, but MJ and ChatGPT are the 2 that I’ve kept the entire time.
I’ve had spurts of thinking about cancelling ChatGPT, but whenever I do, they end up releasing something that I need to cover. Haha, maybe I should just start thinking about it to hurry GPT5 along!
So the wireheading is coming along nicely.....🤣
Can it recreate people who have successfully learned to eat with their mouths closed as well, or does it only work for Mainlanders?
7:15 yes more or less those was exact trained images
Tim, this is extraordinary. I'd love to get your analysis on the Kate Middleton video, where she appears in a garden, seated on a bench, talking about her recent alleged illness. I genuinely think there is public interest component to the video, your professional opinion would be valuable. Regardless, thanks for your outstanding content.
Yeah, I watched that a bunch of times as the speculation was mounting. To be honest, I think it's.a real video. It does look a little "off" but mostly because of that super diffused background. There's also this think that tends to happen when we see the framerate conversion from BBC footage (25fps) to US (29.97fps).
I'll admit though, when people started talking about it, I had to go back and watch it a few times. This is not going to be the first time this will happen this year, I'm sure!
@@TheoreticallyMedia Thanks Tim, broadcast professional myself and just wasn't quite sure what to make of it. Appreciate your reply, thanks again.
You mentioned facefusion and another face swapper that I'm not familiar with. What was it? "Groop"?
Roop. It’s still pretty good!
2:35 - Don't be suspicious, don't be suspicious. xP
That faceswap can't be the standard ones like FaceSwapLive, FaceFusion, and Roop. This is something more impressive than all of that. Everything holds up perfectly. Not a single glitch. Impressive. I wonder if they'll release it. I also wonder if it's just a hyper-specific model for exactly his face which can't generalize.
Aw man your vids are my bread and butter!
Haha. I’ll take that with a side of bacon and some coffee!
My philosophy is: 'If it’s not released to the public, then it’s not ready for discussion.'"
I totally get what you're saying-- but at the same time, I like to keep track of what is possible, and as an eye on what will arrive in the span of three to six months.
For example, when I covered the Adobe MAX sneaks and we all saw Video Inpainting. I can point (somewhere in my archives) to the whitepaper on it, months beforehand.
Hyping it is one thing, which I try not to do-- tracking it is another.
That wasn't meant for you. 😃 I get upset with these people when they brag about how good their AI is but then don't release it.
Imagine a world where you think you are a#10. Every mirror,and camera you look at is actually a digital display that spits out a #10 image of you.
7:02 is that midjourney's webpage version?
it is! Accessible now if you've you've generated over 1000 images. Should be fully open any day now!
Good night.
How do you ensure that a buyer, after buying a photo of you, doesn't sell it as if it were that buyer's photo?
What security does shutterstock give you so that this doesn't happen to you?
Thanks
did I talk about Shutterstock in this video? I don't recall that. Anyhow, as far as that would be concerned, technically you'd have a licence. I mean, no one can stop someone from screengrabbing and posting it elsewhere, but that's just an overall internet problem.
@@TheoreticallyMedia I only write in this video because is the newest one in your channel. Thanks
I think it's time for an AI art competition where people compete to have their idea framed on your wall :)
See, this is why you're the king (and queen!) of competitions!
Always next level content mate!
Much appreciated!
Also if you can’t find a painting, maybe you can just put up your favorite midjourney pictures that you have ever generated 😊
That’s a really great idea! /imagine “Live, Laugh, Love”
Haha
@@TheoreticallyMedia 😆
What's the best ai video face swap available? Doesn't matter if free or not
He's speaking in mandarin with a sort of southern accent, so is this tech coming out of mainland china? coz I know Alibaba has been working on sth similar
Thank you
Spline is already doing 3D with text prompts. So MidJ has to catchup to that
Perfect Video, thanks! Synthesia is good expensive, "aua", like we say in germany 😅
oh, I haven't seen pricing yet! I'm sure considering the amount they're spending on training that it has to be more than the cost of a Milka bar! (haha, I remember those from being a kid in Germany!)
@@TheoreticallyMedia 😅
Which faceswap app is this ?
We don’t know, he doesn’t say other than it is something he trained himself. I’m keeping an eye out for more info!
8:03
😆 🤣 😂
Too funny!!!!
Haha still laughing re Santa joke !!! 😂
He's speaking Mandarin. If you watch movies it can help to think Crouching Tiger Hidden Dragon (Mandarin) vs Kung Fu Hustle (Cantonese)
Also, Cantonese being a southeastern dialect has a certain stacatto that reminds me of Vietnamese, whereas Mandarin is generally more soft and lyrical
Ahhh- That is a super helpful tip! And EXCELLENT references!
Has anyone tried the Faceswap on Nic Cage and John Travolta?
both their significant others have with photos
Faccccce Offffff! Not the greatest of John Woo films, but still a classic of modern cinema!
Didnt “real illusion “ make this like 15 years ago?
Check out the hands
Haha, no handswapper yet!
Emotions on "machines", disturb me some what.
Yeah, in that one example where she transitions between emotions, there’s something…disturbing about that.
Tim is really a Scandinavian supermodel
I've been faceswapping myself to this ugly mug for years! Busted!
Him apologizing for studio B when we can see only two walls is hilarious.
Haha, Studio A has a much cooler backdrop. Studio B is just...boring.
Gonna try to hit up some random shops today to see if I can spice it up a little!
An orbit legend... Ah !
that face swap was impressive lol
very neat
Thank you!!
Why is no one talking about the fact that you can see clearly see that the guy on the "original" view is supposed to be wearing glasses but they are erased by AI. They pop in and out many times.
What are the best face swappers out there?
Before your print that image of the detective, I hope you Vary (Region) his left hand.
that first faceswap should be fake, he did some filter on the first one, and the second onf is the deepfake .as first is perfect but the second has some movements that looks like not live camera . its edited version
For two years already I see everywhere that ai art will die within a year. Looks like for many more years to come I'll be reading this...
Nah, it’ll just be implemented into many artistic processes. Most will use it, some will not. It’s basically like predicting Photoshop v1 will die in a year. Obviously didn’t happen, but there is still a number of artists who still work solely in analog.
Mandarin baby. All the "zh" sounds.
Thank you! Chinese dialects (and in general the language as a whole) is a mystery to me.
I’m below average in Japanese, and around “struggling” with Korean- but for Chinese, I’m 100% google translate!
@@TheoreticallyMedia Hehe I was the same. Only when I started visiting Taiwan and China I started to learn. Before that Japanese was much more interesting. I can kinda survive in Japan but can't have a conversation. I always forget all my Korean and have to learn it again whenever I go back (-:
Is there a legitimate use case for face swapping? I'm thinking there isn't.
Definitely Mandarin. Pretty cool stuff.
wow , reality sometimes surpasses fiction !! glad to be alive on this times
It's a pretty wild time for sure!
And fiction is often the inspiration for reality, especially with technology.
How to make a substitution of a deepfake of a face with a beard?
That, from what I’ve seen, is basically a mess right now. You can go the other direction (putting a beard on) but a digital shave isn’t doable without extensive manual VFX work.
Man HANDS
he speaking mandarin. YiLong Ma got famous for that.
I spent the first half of this video trying to figure out if you were completely AI generated.
Haha, if I were, I’d probably say “um” a lot less! That’s how you’ll know: when I stop messing up and actually sound fully professional!
Faceswap didn't account for the man hands.
That's what I keep saying to the doubters!
@@TheoreticallyMedia it's a semi-fake fake. I looked at it frame by frame and these are two videos of real persons mimicking each other but then they ran something like insightface over the guy in the small picture so that is face match more closely the girls face. that is why you can see the guy's glasses popping in and out all the time, same thing happen in roop when you swap persons with glasses, the model is too low resolution for glasses. it's a cool video but the girl is real, not AI. a lot of manual work clearly went into this video.
I've tried to reproduce his results with existing free tools. I got close, but not with that high res. Maybe it is a new tool. Judge by yourself: ua-cam.com/video/afEI4SdPvEA/v-deo.html And sorry, no English today :) just turn on automatic translation
i can completely understand your english, thank you for giving us the link to your channel
we can finally be as hot as our delusions would have us believe we are!! :O
listen...
Why does that guy have long hair? Or is she being converted?
He’s wearing a wig. I once saw a guy do it with a towel on his head! It was pretty funny, but it worked!
1:38 min Never in real time. Here you can see his glasses melting into his face. And I don't think that's his face either.
For the first one, it's likely a sh*tpost from some Chinese guy copying actions of a poor video to gain views. Happens a lot
That’s interesting. Like, he reverse engineered it? I’m not 100% on that, but if so- kudos to him!
the things possible ... haha ... scary haha xD
'') VR live with multiplayer ....
looks like a real BOY!
Synethesia still sounding too synthetic.
As for MidJourney, how long till they can make me tea? Earl Grey, hot!
🎯 Key Takeaways for quick navigation:
😮 A mind-blowing real-time face swap technology that convincingly tracks facial movements and expressions.
🤖 Synthesia's new AI avatars with improved lip synchronization and emotional expressions.
🔮 Midjourney's 12-month roadmap focuses on video, 3D scenes, real-time rendering, and an interactive world simulator.
🎨 Midjourney's new "Style Random" feature allows randomizing image styles, offering diverse artistic looks.
🎥 Morph Studios, a new AI video generator with a node-based UI and consistent character styles.
📹 Nim Video, another AI video platform with features like image-to-video, video restyling, and open-source models.
🌐 Both Morph Studios and Nim Video offer beta access and promise exciting AI video generation capabilities.
💻 The ever-evolving landscape of AI tools for image, video, and 3D content creation.
Made with HARPA AI
The face swap looks fake, after you pointed out that the eyeliner was off, it seems off in a lot of it and almost like she is watching something off camera.
I keep going back and forth. The one that keeps tripping me up is the hands. Like, on the swap, if it is faked-- does this girl have big meaty hands?
Either way, if it IS faked, it is a really really impressive mimicry. Like, that alone is worth a watch!
@@TheoreticallyMedia True the hands are meaty. But if you also look at the background on his BG, its shooting off on the top right edge and hers isn't, plus the hair is falling differently on the face on the right. I think its mimicry, But I agree, good job if thats the case.
@@TheoreticallyMedia translate face swap to Russian and see what they are running. Plenty of videos in UA-cam and communities on telegram. They also have whole writeups on it and systems in Chinese. You can get them from their repository. But harder. Really, the best, is when th Russians grab the Chinese faceswap or AI and redo it to English or Russian. The issue with Chinese is that sometimes you need to log in with their systems or go through a verification process to get the software and if you don't have a WeChat, it's a pain. It's a whole world that US does not get to see. Very advanced avatars and swaps.
i do not think its real, as you point out things dont line up, but im talking head shoulder neck, as in his should and torso don't line up and her head neck is pointing down further. to me its two people and its been rehearsed both filmed and synced, probably to drive up investors on the real program
I’ve heard that speculated, and if that is the case, I have to say: that makes it worth the watch as well! Because if so, that is SUPER impressive mimicry! Like, that should be an Olympic game!
(Seriously, I would watch that!)
@@TheoreticallyMedia oh man you need to youtube fun house mirror mimic.
haunted house people doing it on the fly with no choreography, there eye to response time to copy is insane
@@TheoreticallyMedia ua-cam.com/video/qW_gKJzJKNw/v-deo.html
Eversince photoshop, they have advanced significantly with AI, enabling deef fakes in everything. I consider most stuff on UA-cam as movies, not real.
5:07 3d has been held back by lack of data = we couldn't find enough copyrighted work to steal yet, like we did with art
I mean…you aren’t wrong.
And it’s possible that as the move toward 3d happens, it’ll resolve a lot of the training data issues, as it is a new workflow that (might) require a new dataset that isn’t as readily available.
OnlyFans, here I come!
Next up: a good Foot Swap!
I want a tool to swap my messy house for a clean one.
Oh wait, they already got that, Apple Vision Pro
scammers new materials for sure
oh, we'll be seeing that for sure!
Mandarin
Thank you! I can speak a bit of terrible Japanese, even worse Korean, but Chinese (and its many dialects) break my brain for some reason.
Okay, so let me just say that this faceswap is real enough and it's Chinese. There's a video "out there somewhere" that shows how this is used in an almost industrial format. China is far ahead in this area.
Yup, got your email! That’s wild, I knew it was a thing in China, I just didn’t realize how big it was!
@@TheoreticallyMedia Yes, and they don't care much about morals and ethics regarding deepfake, etc., which is what is holding back the West's (USA) development in the area. From what I've heard. That's why they're way ahead of us.
Which tool is the one with the face swap in Elon Musk?
That one was called CamerAI. Points to being clever with that one!
@@TheoreticallyMedia Thank you 🙂
I think the face swap was done by a guy using a wig to illustrate the software's capabilities.
Yeah right, $22 a month but only 120 mins of avatar video, A YEAR, 😂 for synthesia.
Ooof. I hadn’t seen pricing on that yet. That’s steep!
@TheoreticallyMedia thank you for the info and the video! I am super interested in the latest ai stuff. This was the first video of yours I came across. You got a sub from me.
I guess my comment got removed 😒
I don’t see it? Maybe? That’s weird. I see this one!
@@TheoreticallyMedia Maybe in your Comments for review section? And maybe because it had a trunk-a in it? UA-cam..... However I can repeat it here without trunk-a...
@@TheoreticallyMedia I wrote: I saw something very similar to that Faceswap several months ago from China. It's a guy who reports all sorts of weird things happening in China - name 'serpentza' and search for "China is now Making Fake Women" uploaded 3 months ago. I'm not saying it's the same app, but it's the same topic.
I have past it again but I don't see it unless I hit "Newest comments". I hate UA-cam
@@bpvideosyd its gone ballistic, lists or references to other channels, it removes, also sarcasm 🤣🤣
As a brit im skrewed
His cheeks.....????? It's a lady....😮😅
Synthesia is cool, but in regards to customer service bots, teachers, mentors, etc, do people really want Smiley McSmilson all the time? Heck, I don't like that in real life. We want genuine emotions. I'm being silly and picky, but anyway. Thanks for the vid!
If you slow down the video, you can tell this is a fake deepfake. Someone spent a lot of time to do this, but looks like old school lip-syncing to me.
I love that we've entered the age of needing to scrutinize a deepfake to see it is a...real fake? Wild times!
nice video. Can you refrain from just casually tossing out Santa murder jokes as a way to keep social norms healthy. thanks.
Fair deal. Santa Murder jokes officially retired from the channel.
No promises about the Easter Bunny though. =)
I'm quiet impressed with the first 2 tools. The last 2 video tools = 🤮 quality.
I think Adobe is on the right track with their AI video tool approach in terms of integration into a formidable tool suite.
Unfortunately, access to that tool suite will come with a $$ price tag.
Nah 1. its chinese 2. it looks maybe a bit too good compare dto big tech versions ( though the video quality is low) its probably fake.
I dont see any benefits in that rather than use it for the worse like identity theft, ruining someones reputation, false propaganda and practially all other illegal stuff. so why????
I mean, valid question. There are some practical and fun uses. On the practical side, you could use it in films for when an actor isn’t available, but their stunt double is.
Or on a more useful level, faceswapping to your own face for Zoom calls when it’s Monday and you just don’t have your game face on.
But, yes- also 100 other bad uses. But, that’s just technology in a nutshell. It’s always a double edged sword.
@@TheoreticallyMedia is this actually already included in the law? whats legal and whats not? I wonder with all those Tom Cruise and Keanu Reeves Deepfakes popping up on social media- how that can be allowed? its identity theft. Is the law behind in this?
How is any of this is a good thing. It’s nothing more than being deceptive to the masses
It seems typical to make yourself a little girl in that industry.
Whatever you did to your audio this video is VERY distracting. It's very warbly-sounding and very synthetic sounding. It's as if you ran it through too many noise-removal tools. Very distracting.
He did say he was in a backup studio. Not super distracting to me, just not as clean as normal
@@CGingerbreadman No of course, I understand that. But something else was done to process the audio which is very distracting.
Thanks for the feedback! Yeah, first day in the B studio, trying to dial it in. It’ll be better for the next video:
That warble is actually Adobe Podcast Enhance, which usually does a pretty good job!
I’ll have it dialed in for the next video! I also haven’t moved my speakers up, so this was laptop speaker audio mixing, haha. Terrible!
@@TheoreticallyMedia Thanks for your response! I'd say if your pushing the Enhance Speech function on their browser based server, try not to exceed 40% on the slider. It tends to introduce a lot of artifacts into the audio. But seriously I love your content, it keeps me up to speed on a lot of what my industry is involved in. 😊
My --sref random result is always --sref random, not like your --sref 12312313213 @ 7:50 ? How do you switch to that? In other parts yours is like mine, just says --sref random
Yeah, I noticed that too. I THINK it takes a bit to refresh. It's a fairly new feature, so I'm sure it's a little buggy right now. I didn't want to fully call it out, since these videos tend to stick around for awhile, and I'll be pointing out a thing they fixed a long time ago.
Checking now, it appears the images I generated earlier seem to have SRef numbers, so I think it just takes a bit.
@@TheoreticallyMedia Thank you! I will wait and see.
7:02 is that midjourney's webpage version?
it is! Accessible now if you've you've generated over 1000 images. Should be fully open any day now!