Disney’s neural rigging elegantly bridges the gap between complex machine learning and intuitive artistic control. It signals a future where theme park immersion is no longer a static show, but a real-time reflection of the guest's own identity.
Deep Dive
Prerequisite Knowledge
- No data available.
Where to go next
- No data available.
Deep Dive
Disney Is Testing Something...Added:
Good evening friends. Bricky here. Going to be doing another watch along. Uh Disney has a research hub YouTube channel where they upload some of the strangest videos I think that you'll ever see. And often I look at these different technologies and I debate is this genius? Is it is it brilliant? like is it is it something that's gonna change the world or just the next step to the next idea or is it terrible? But I believe this website has to do with some sort of funding they get. So tonight what we're going to do is we're going to get caught up. We're going to watch their latest two videos and try to figure out exactly what Disney is building. Um because when you look at these research videos, it's kind of like reading half of a story. You're just like, I don't know. I guess they're going to do something cool here. So, if you want to chime in, give me your theories, of course. I always love to hear those. So, let me say hi to the chat and we'll jump into our first video, which is um cross attention neural face rigging with variable local control.
Rolls right off the tongue, baby. Rolls right off the tongue. What's going on everybody? How you doing? Let me just get a couple things set here and uh we'll jump in.
Shasha, what's up bud? Good to see you.
Star Wars Mike, how you doing, bud? Good to see you. Hope your YouTube adventure is going fun.
Let's see here.
Did Shashana, my sound guy? All right, I got my m my mic check from Shashana.
Volume sounds good. Are we ready to jump into this? So, here's the thing. Like, I know a lot about design and architecture and, you know, how theme parks work. Um, a lot of this stuff it's just kind of like, okay, I think I know what's happening and then you have to kind of make a guess. So, obviously, I want to hear your guesses tonight. Let's just jump in. Let's get it going. Right, Richo? It was a mouthful. Let's Let's just jump in. We'll we'll take lots of breaks. Uh, feel free to holler at me if you have a theory and uh, you know, hold on. Let me just get rid of this real quick. There we go. And, uh, we can get going.
All right, let's check this out. And if anybody super chats this evening, the lightning lane to my brain, I think I have a new effect figured out. So, we'll figure that out if we get going. If not, no worries. All right, let's jump in and see what we're looking at here.
>> We present Kendrick, a novel crossension neural face rig technique with variable local.
>> You guys can hear that, right? My camera's in front of my metering tool.
Let me give me just one second here. I want to make sure you guys can hear that when I'm playing it.
>> Control for animated characters and captured subjects.
>> A very intuitive and direct way for artists to sculpt.
>> Okay, so what I we'll jump in. We'll see where this is going.
>> The facial expression is to manipulate handles on the face. Some existing methods like the shape transformer by Shantron at all support this rigging principle to some extent. By the way, this guy is just like very charismatic, you know, just has that big radio personality that you love so much. So, I I what I think we're looking at here, cuz we're used to seeing all those little like action points, right? Like when you see the guy with the ping pong balls all over him when they're filming it. Oh, James Brown. Let's see if this works. My homie, the only Brad prom is James Brown. Okay.
Okay. See if this works. See if this works. Save the clip more. Okay. Come over. There's so many easier ways to do this, but did it Why is it not working? Hold on. Hold on.
Hold on. Hold on. Don't give up yet.
Don't give up yet.
No. Oh, there it is. There it is. Don't give up yet. Don't give up yet.
James, it's green. Okay, so how do we do this? If it's green, what happens if we do this?
Okay, so it's there. It's there. James Brown is there. Okay, let's get rid of this. See, I can just kind of like float this little guy wherever I want to.
There we go.
There.
Oh, no. No, no, no, no, no. Don't do that. Don't do that. Dude, don't do that. Do that. Do that.
Is this working? Dude, this if you guys knew if you knew how how wild this is. This is like this is this is what I'm doing is the low end of technology.
But hey, I wanted to be able to give a shout out. James Brown, thank you for testing with me. I appreciate it. Okay.
To the left. Okay. Let me come back over here. See, I can't do it live. I'm like a weatherman when I do this. All right.
All right. That's good enough. Okay.
All right. Let's get it going. The shape transformer, however, has no sense of locality. If you edit one single handle, the entire face can and will change.
>> Oh, did you guys get that?
Did you guys get that? So, when they're moving the little yellow thing around, as it gets close to all of the little blue nodes, they have the whole face mapped out to where if like there's like a force like gravity, right? So, when it goes around, it's picking up the tension of everything that's happening. So therefore, in the world of animation, this would eliminate a lot of steps. Or if you're doing motion tracking, it would like make like if I had these dots on my face and I'm moving around doing my voice over, then while I'm doing my voice over and I'm acting in the booth, I'm also animating my character. Dude, sick. Sick.
Yeah, the chat's over my face. You don't need to see me. You just need to see James Brown's there. All right, I'll move it then. Jeez, everybody's a critic. But if I put it over here, it's over my blue friend's face.
And now I got to bring that back. Right, we we'll figure this out. We'll figure this out. There you go. I'm wearing it like a hat. All right, now that that needs to move. It's on my dude.
Let's put it here. There we go. There we go.
All right, let's jump back into our dude.
Yeah, it's like a mesh. Jay Harris, there you go. You're on the same page with me.
>> In contrast, methods like animat subdivide the face into several local blend shapes that are then edited directly by the user. In the case, >> oh, so when they move this wire frame, it basically moves the mesh JS that's a great term all around it. any it's a single blend shape per original muscle.
The downside of such a method is that the locality is hardcoded during construction of the rig. Whatever he just said the downside is it >> changing the level of editing detail during runtime is not ang orange.
Oh god, now the chat's moving. I didn't anticipate the chat moving while I was trying to The chat keeps moving while I'm trying to do the screenshot.
This is so ghetto.
Okay, come over here. Uhhuh.
Yeah.
There's no way anybody is amused by this as much as this as I am. And so now I can stack it.
And I can also do this.
I'm done here.
Oh, I'm gonna have so much fun with this layer.
Thanks, Hank, for for helping me out.
and I appreciate it playing with my new toy that I made.
Okay, back into our science experiment.
>> Possible.
We propose a method that allows the artist to edit the shape through handles with a custom influence region. This region can be small and everything remains unchanged, medium sized, and we change most of the mouth.
or span almost entire face and we get global correlations from a single handle.
>> Dude, this is so cool.
Michael Mark Hopper green. Okay, we can figure this out. We can figure this out.
Thank you, Michael. All right, so what's that look like to you guys?
Okay, you can see the text. What happens when I go inverse?
Okay, pink. We'll do it. We'll let it go.
So, just put this over here in the pile.
There we go. But we also will show James's name.
I'm here, guys. Trust me. I'm I'm here somewhere tonight.
I'm here. Don't worry. You don't need to see me. You know what I'll look like.
My camera's in the way.
Oh, did not mean to do that. Sorry.
Sorry if I'm playing with Oh, come on, dude.
Hold on real quick.
There you go. Now you can see my eyes.
I'm still in here. I'm still in in here.
All right. So, this is what I'm seeing.
Dude, I don't know, Steve. If you're just joining the stream, you don't know what's going on. Neither do I. All right. So, what I'm seeing here in this example is the dude on the left that looks more human showing us what could happen. tell me if I'm right or if I'm wrong, but it's showing me like so let's say that I'm the model on the left and they can take my mesh my face and then articulate what I do for the animated counterpart of me.
So again, my theory, if we're looking at this as like an animation thing, if I was a voice actor and I'm in my booth and I'm reading all my lines and I'm acting everything out, they're essentially getting the illustration of my face and then that gets converted over to the character, but because of this flexible influence radiance, they have less dots, more animation. Am I seeing this right?
Does that feel like the right thing?
But it's uh Star Wars Mike. It's not just a motion capture demonstration.
It's an advancement. It's like a new technology that they they they're showing in a research paper.
>> We're not sure how we can design new shapes using Canri. First, >> this looks scary.
>> We add a set of handles in green and select some handles in yellow to edit.
The color map on the mesh visualizes the influence area of these handles. Now we move these handles up to achieve a snull face. Note how the upper part of the face doesn't move.
>> So if you think about this from an animation perspective, if you made a face and you put this software over it and you had all your bin points, like instead of actually like physically drawing on it, you could just do like command scripts where you could just be like, "Okay, I want this to go up and go down." put in your your your uh amount of frames you want and you could basically drag and drop animation all around the character.
Oh, don't worry J. Harris, we're we're going to get to how this fits into the theme parks.
>> Sorry, that was rude.
>> Now, we activate new handles around the left eye and repeat this. Now, only the part in this influence region is changed. All other parts of the face remain still. Oh, so you can just do Pretty cool. I'm not going to lie.
It is kind of scary, right? Magic space.
All right, magic space. We're gonna do this one fast. Now you guys are chatting and it's moving on me. Magic space.
Magic. Don't move. Don't move. All right, I got magic space.
Did I get it?
All right.
Okay. Get Magic Space. Thank you, Magic Space. Appreciate you, bud. Okay. Come over here. Put Magic Space in. Boom.
Hero image.
And now put Magic Space over in the pile.
Oh yeah. Yeah. You can see the community hanging out with me. This is sick. Love this.
Okay, I'll figure it out. I'll figure it out.
This is test run. This is test run.
One downside of the method suffice that in order to edit a sequence captured with motion capture here on the right, we first have to invert it into the rig space. This inversion is simply done by driving the handles from the ground truth mesh. And while the reconstruction is fairly accurate with an error usually below 4 mm, it's still lossy.
>> I have no idea what he just said, but I'm watching I'm watching the reconstruction, the error, and the ground truth. I'm trying to figure out what exactly the error is. I think these red heat indicators are showing us where it's like off up to like 4 millimeters.
Yeah, we're we're we're building the the $20 van.
Magiccape, thanks for calling that out, Star Wars Mike.
And if you ever were like, I wonder if he's really dyslexic. Well, I just called my man magic space, not space.
Welcome to my brain, baby.
Can't tell the difference between space or escape. But I'm going to try to figure out how Disney's making a robot man in a theme park.
We would like to avoid this freaking version further.
Oh, like this. It'll be better when it's just the office. See? See? It'll make sense when we're there. Everybody's just kind of hanging out, chilling.
>> Okay.
>> Assuming you have a shape in the rig space already. Another downside is that iterative non-destructive edits are difficult. If the artist wants to close the eyes but selects only a subset of handlers, these constraints can fight and produce undesirable shapes like here on the eyebrow. So still more an eyeopen eyebrow raised shape. For all these tasks, we need to be able to iteratively layer edits non-destructively.
With iterative editing, we can simply add a new layer and move new handlers without fighting constraints.
How do we achieve that? Kenrik allows the user to specify arbitrary handles on the >> canrich. That's the name of the software that we're looking at that they created.
Canri cross >> face and move them around. These handles define the keys and values of >> delta landmark.
Okay.
>> Attention system. We want to reconstruct.
>> Dude, with all my chats on here, this is becoming like the white trash Disney version of beautiful mind.
Oh, this is great.
Park mind. The technology is both fascinating and terrifying. It gets in the wrong hands. Yeah, I know. That's that's with all this stuff, right? It's like Superman's dad said, with great power comes great responsibility.
>> Full shape. Hence, we use the vertex positions of the canonical mesh as query using >> Oh my god, it's getting wild in here, guys. It's getting wild in here.
Too much math.
Jay Harris, I get that it's just an engineering demo, but did it have to be so death mask like? I know. I love I love that it's like, hey, let's just animate a cadaavver's face. Okay, so let's see if we can follow this.
So there are the handles up top left, bottom right is the mesh.
So the data is coming over.
The data is coming over and connecting at the attention matrix.
Softmax rejoins the value.
That's where the data collides and creates the output of the predicted shape. All right, I'm following it.
>> A single cross attention layer. We then reconstruct the edited shape as a displacement to the canonical mesh.
Right now however nothing limits the network from learning global correlations between the handles and vertices. We therefore introduce an >> new variable >> explicit mask to the attention matrix that controls the exact region of which handlers can influence which vertices.
This mask can be changed at will even at inference time to support iterative edit.
>> Oh, hold on. I'm in the way. I I I'll be back.
>> First introduce the base shape and next the handle edits are defined no relative to the base shape. And finally, the output shape is also predicted relative to the base pose.
Nerd. This dude's a nerd. Look at this.
I failed physics. I didn't even take physics. I'm just going to figure out how to make a robot man. I'm just going to skip all that stuff and just make a robot man. All right, I think I'm I think I'm following along here.
>> Let's look.
>> All right, so let's look at some car can rig applications. This is where it gets fun. Half of these videos you're like, huh? And then you get to be like, okay, now I see where it's going.
>> Some applications of our model. The first scenario is an A to B test. We're starting from a >> Okay. So, A is the base. B is the production or no, A is the base. The prediction's in the middle, but B is the target of where they want to go. You got it, J.
Harris makes total sense, right? Oh, King Thanos. I like that. I kind of feel like it look Oh, I was thinking Mr. Incredible, but I could see a little King Thanos.
O, how well can the model reconstruct the target shape be on the right?
And here you can see this that we match the target shape pretty well for various number of different A and B pairs.
Here are the results for a second character.
Here's the murder fuel that's on the cover of the video. I just like scrim through. I go, "Yeah, that looks creepy.
That might get people's attention. Let's put that on the cover." Yeah, Mr. Incredible. Right, Carl? My brain hurts a little bit. This is the fun of it, baby. We're doing a mystery together >> for various >> Dude, I would love for the dudes over at the research lab to do a reaction video to my reaction video and just like this guy is so dumb.
He doesn't even know the difference between space and scrape an A and B pairs with just the sparse SL.
>> Dude, so much chili on his face. marks.
We can match the shapes pretty well.
>> Okay.
Performance.
>> Yes, we show some applications on entire performance.
>> Looks like Jack Jack, right? That's why the other guy made me think of Mr. Incredible. Totally Jack Jackances.
We start with an input performance from motion capture. In a typical visual effects scenario, we don't want this exact performance. But >> pretty se that creepy face in the thumbnail is exactly why I joined.
>> Want to do some small tweaks to make some expressions read more for example.
So let's stop at this frame.
>> Okay.
>> Note that we use this as a base shape.
So we don't need a lossy wick inversion.
>> We can directly add handles and move them to the desired expression.
This edit is then smoothly integrated into the performance. Watch how the mouth shape is changed for only a single expression. All the other parts of the face are left unchanged.
Oh my god. So, think about this. Let's go back to my voice actor uh example for a second.
So, I'm acting doing my voice acting.
They're capturing out my motion. So, while I'm doing the vocal, I'm also basically editing my face. But let's say that like I blink or I have like a weird tick or whatever that doesn't fit what the character is doing. They can go back into my capture and just s like isolate this eye and like keep the eye wide open or open them even wider if they wanted to be more drastic. Or I bet if my eyes like shift off at my script, they could digitally have somebody that keeps my eyes looking over at the character that I'm talking to. Holy smokes. This is cool.
Don't ever take away my chats. Ever take away my chats.
I don't like that little gray bar on the side of the screen. We'll have to get rid of that. We'll figure this out, guys. We'll figure this out. We're just having fun. They're trying to figure out this. I'm trying to figure out how to make my chats pop up. Everybody's got a project.
We can take this a step further by editing every single frame.
This guy's really freaking me out.
Somebody said he sounds like Stitch.
I've never seen Stitch, but I'll take your word for it.
>> Given an input performance, let's replace the mouth completely by a different per >> Oh, yeah. It's doing exactly what I predicted. See how the eyes on the right are doing something different than the input >> performance from the right. We use the same approach as before and move the handles around the mouth of the input performance to the target dialogue for every frame. The result is a merged sequence with the mouth from the right sequence and the rest of the frame unchanged from the left sequence. To allow artists to directly interact with the rig, we integrated the network into Blender.
>> Oh boy, Swede. Sup?
>> First, we show the different influence regions when hovering the mouse around.
Then we apply edits at the or region and around the eyes.
Wow, this is so cool.
>> Next, we update the base pose since we want to increase the radius and do more global edits.
This allows us to easily open the mouth for the final edit.
Next, we show how the proposed rig can be integrated with other rigging methods. First, we distribute a dense set of handles over the face and add our jawbone. Then, we can use linear blend skinning to drive the handles. This allows large edits that still show actor specific nonlinear motion through the network.
>> This is insane. Similarly, we can use curved sliders mapped to local linear blend shapes to drive the handles.
This result can be used as the base shape for iterative editing. By adding new handles, we can realize additional local edits.
Okay, so hear me out. Hear me out.
Here's my thought.
Here's my thought. Right.
You hold on. Let me just get my children over here.
Let me just get my my beautiful collection where it belongs.
Okay.
Okay. All right. So, hear me out.
You go into a theme park ride, say the Haunted Mansion, there's a sensor on your Doom Buggy that tracks the nodes on your face. Real simple. It just can do like a cross across your face, right? And it can kind of just put a couple of simple spots where you're at. So then when you come by like um you know the ending scene where the hitchhiking ghost is or the um graveyard singers, it's your you know in the in the hitchhiking ghost. Whatever you're doing in your face, the ghost could be doing back to you because it's tracking you. But if it could also go by and do a quick scan of your face, it could also basically make your face one of the singers in the graveyard. So as you and your friend come by in your cart, it's your dead face singing back at you because it could quickly scan it and digitize it.
That's the type of thing that I'm seeing this technology is going. And I think that like if it's a temporary photo that is taken and not saved, I think that that's well within the the law. That could be insane, dude.
That's awesome. And you could see how from a illustration perspective how that could skip so many steps.
Yeah. Make us into ghost, Steve. You're following along. You got it.
Glam Blue Eye says, "I'm trying to figure out how to afford my next Disney trip." Well, all you got to do is just figure out how to make this software work and then you're rich like this guy.
Michael failed physics, but now I get it. Wild. All right, that was our first video.
That would be scary for the other guests. Well, I mean, it would be personalized to you. The only way that this like animating your face stuff works is in one-on-one situations. So, the way that I'm reading this as far as a theme park application is is that it has to be able to capture you and then in a vignette where only you are watching. So, the reason why I use the Haunted Mansion as an example versus like Pirates of the Caribbean in the Doom Buggy, it's just two people because the Omnim Mover could move around. It would just be who's in your doom buggy looking back at you. So, the next cart, it won't happen until they're there. And I think that this type of augmented reality stuff has got to be the next level of like theme park enhancement, right? Like it it can't be too far away.
Escape for the day. I mean, they already have our access to our face when we walk in the park. 100%. All of our faces are recorded.
In fact, they know when you show up on property.
I like this stuff. I've done videos about it before and other people don't like it. Kind of freaks them out. Kind of creeps them out. I think that this is cool. I mean, if the technology is there, let's use it, dude. Richo Max Hedrum shout out. Wild.
All right, I have one more. You guys want to watch another one? Let's do another one.
Just let me let me get my children out of the way here. We're going to go with like a top look.
Okay. Magic space. Oh, layers bricky. Layers bricky. If you guys only knew how low tech what I'm doing is. Okay.
And then this one is There we go.
Boom. Look at that. It's beautiful.
It's beautiful. All right, let's go over to the next video that I have.
This is another video from the exact same website.
More technology.
Sounds like maybe I accidentally downloaded the non-English version. We could probably follow along, right? We got this. We got this. Let's do this.
We know how We know how to do chicken in in whatever or drunk.
This might I understood chicken and drunk. This might be where this gets complicated.
Okay, let me see. Is there a way to switch the language? Is it Russian? Is that what it is?
What is this word? Bricky has zero.
I can't read this word.
Ability to judge. Well, Bricky has zero ability to judge. Well, that's not nice.
How dare you? How dare you say I can't judge. Well, look, look at how well I'm doing with all my chats on the screen. How dare you?
Thanks, Casey. I appreciate it. Please go into my collection of treasures.
Thank you. Appreciate it. Appreciate the support always.
Okay, we gota There's got to be a way to switch this.
Give me one second.
All right.
Okay. Okay. So now we are doing the dental disintilent style and content in motion with residual quantized representations.
Nailed it. Let's do this.
Let's do this.
>> We present VQ style.
>> Oh, AI voice.
Oh, right.
>> We present VQ style, a novel method to disentangle content and style of human motion that enables multiple applications, including style transfer.
Charar is Oasis going in the Hall of Fame. That is amazing.
Love Oasis.
All right. All right. Charger.
Charger. Friend, not foe. Thank you, buddy. I appreciate it.
When I bring these back up, don't worry.
You're in the collection.
You're one of my infinity stones.
Okay, let's figure this out.
We train a residual VQVE for motion reconstruction.
>> Mhm.
>> To disentangle style from content, a style-based contrastive loss is applied to the residual code books.
We further add a mutual information loss to discourage any style information being present in the first codebook.
>> Okay.
>> As a result, the first codebook contains all of the content features and the residual code books contain the style features. Of course, >> by swapping style codes from a style reference clip, we can perform style transfer.
>> Okay. Roadrunner.
Pink is style clip. Blue is content clip.
The third, the multicolor is the style transfer result. So this is what I'm assuming.
They create styles and then that style can be applied to a content clip which is kind of like a rough rough grading of your character walking through the space. And so you go, "Oh, the character who's walking from point A to B, I want to run the Roadrunner style on it." And then it does the Roadrunner. You think I got it?
Yeah. Look at that. Roadrunner. Row runner. Oh, the pigeon toad.
Oh, Kevin, you're one of my infinity stones. Oh, stop when you guys chat.
Okay. Okay, Kevin.
Saved my clipboard. Put you in my collection. One of my stones.
Didn't work. Didn't work.
The Kevin Cruz stone is one of the most elusive stones for my infinity collection.
Uhhuh. Uhhuh.
Why is it doing color balance? You're supposed to be There it is.
Ah, yes.
Right there. one of my beautiful infinity stones. Thanks, Kevin.
Appreciate you, bud. All right, so pigeon toad.
Think about it. And I know I say that very, very Kentucky. Pigeon Toad. Hey, man. You pigeon toad. All right. So, think about it. You have an attribute that you can apply to a simple a base illustration. So, you can actually run the characters characteristics as like a filter over your animation.
All right, I'm feeling it. I'm feeling it. Because when we look at the content clip, it's literally just moving the character around the environment. But when we look at the style clip, oh, dude, that's wild. Look at this. The character just got its hands down. But the style clip is making it do that thing.
Drunk guy at bachelor party.
Oh my god, this is wild because the blue character is just basically like moving through the space. So, almost think about it if you had it like an Xbox or a PlayStation and you just kind of move the character through the room and then you applied the style of like happy or pigeon toad or chicken or drunk or angry and it would then add that to it. Oh my god, I love when the chat's just so far away from what I'm doing. Where do you think Where do you anticipate they run the treated water from for Avatar Ride and other water features? Well, I did a what a question. I did a video I I love my community because this is the only space where somebody actually would ask something so nerdy and I would actually care so much about that question.
Thanks, Vegas Raider. Uh so I did a video explaining where they're putting Avatar. There is an underground tunnel that connects Disneyland to DCA. You can go see that on my main channel.
So, the underground tunnel says that where the parking structures at >> that they're Thank you, Lynn, for subscribing.
Where the parking uh hub is, that's a part of the park where they're able to go subterranean. I believe that what they will end up doing is they will put a water reservoir in that subterranean space that will sit out back of it. Just like how uh Tiana's Bayou, formerly Splash Mountain, it has its own water reservoir. So does uh pirates. So when they in Small World, so when they work on those attractions, there's always a reservoir where the water can go. The only one that the only closed body of water cuz there's two different types of water systems. Dude, I can get as nerdy as these these research hub guys.
There's the dark water system where it's all the water that's connected that goes from the Rivers of America over to the Jungle Cruise. And then there's the clear water system. So, like the submarine voyage, that doesn't have its own backup water, but like when Grizzly Peak gets drained, it goes over into the Pacific Warf, but World of Color doesn't have a backup reserve cuz it's way too big. Um, or no, does it have a backup reserve?
I don't think that it does. I think they actually have to manually pump that and refill it now that I think about it. So anyways, to answer your question, that is a subterranean space.
They will put a water reserve somewhere that's underneath the showroom or behind the showroom where they can take the water in and out. Fantastic question.
Love that. Now, let me get back to my Barbie dolls.
Dude, look at the sweep. Lean back. Lean back. My robot don't dance. It just pulls up its pants and lean back. Oh yeah, zombie. Sick zombie. Okay, the blue dude's basically doing nothing.
Walking stick, right?
>> We also train our model on a non-locomotion data set and compare with Gen Mo style.
>> Okay, doing something else here.
Might be as nerdy, but voice is not as boring. I mean, I find that the robot the AI lady is way easier than the real nerd dude.
the way of dark water.
By the way, in case you were wondering, here are all my infinity stones, okay?
And don't try to take them from me. This all I care about is my infinity stones.
Okay? We have a style clip. We have a content clip.
Ours Jinmo style. Rockum Jinmo style. So the style clip as we just saw was the pink one. The blue is the content.
So the hours classify predictable label. All right, let's just see what the hell's going on here.
Old. Oh, old zombie.
Oh, I see what they're doing.
They're adding depressed and drunk to get the Jinmo style, which I'm assuming the Jinmo style is a combination of multiples.
Rockcom Jinmo style.
We achieve a higher score in style preservation accuracy.
>> Good for you.
I'm not going to lie, this one I'm having a hard time with.
I don't really know what I'm looking for here.
>> However, in some cases, both methods fail. You know what, robot lady, that's okay. We don't always win. Like me, for example, when I first started doing this, I didn't have all these infinity stones. I had to go out and capture them one at a time. So, don't worry, robot lady. We all fail sometimes. But don't worry. Somebody will prompt you. You'll figure it out >> by only decoding the cont. Let me put my stones up.
>> Codes. We can perform content extraction.
>> All right. I like this version better.
Robot lady. Very judgy Star Wars mic.
Rockum Jimbo style. Yeah. Sorcerers Workshops over there rocking out.
Yeah. Look at that.
Got a pee.
Yeah.
By varying the strength of the style code, we can perform style interpolation.
>> The right hand from style model is more accurate in some predicted models than others.
All the pins and all the stones. Hey, I got a great pen story to tell you guys.
So, remember last time I was live, someone was like, "Bricky, where's the kingpin?" I'm like, "Oh, let me go get it." And I went over to my bag and I showed you the kingpin and then I told you the story about how I pricricked my finger. Well, um, I set the pen down on my desk and when I went to the park yesterday to film on the way there, I was like, "Oh, I left the king pin at home." So, the way that it works is you got to trade a p, you know, I got to have the king pin to trade for the next pen. So, I'm like, damn, I want to be stuck at 28. And then I'll give a shout out to this gentleman.
And then this shining knight walked into my space, Jinmo style. His name is Jonathan English, my homie, my savior.
He walked up to me. I was over by Roger Rabbit. I was filming some stuff and he goes, "Hey, Bricky, what's going on?"
Blah blah blah. And we were talking and he asked me about the pen. I'm like, "Yeah, I left it at home." And Jonathan traded for this pen right here.
Sight unseen. He gave me his address.
And the next time Beth is mailing out our orders, I'm gonna mail this to him because he gave me a Donald Duck pen, which was pin number 29. Then later on, I traded for pin 30. I'm now up to 30 pins and all of these infinity stones. I'm basically rich. So, Jonathan English, thank you so much for trusting to do a pin trade site unseen. You will get this pen and I'll throw a couple patches in there for you as well. All right, let's get back to the science, guys. We got to figure out the science.
>> What's up, Jeff Pike? Ground truth 1.0.
What the hell is happening here?
I don't know. They're losing me on this one.
See, they have to show >> instead of adding the style, >> they have to show this research, right?
Like because you, you know, to get grants and stuff.
>> I'm pin trustworthy, right, John? Wasn't that nice?
>> Residuals, we can subtract them to apply the inverse of a style.
>> Mhm. Wide legs.
Hold on. What did she just say? Sorry. I know I'm the only one that seems to care about this, but I'm in now. I'm in too deep.
>> Naively stitching two motion clips creates a discontinuity in motion.
>> By stitching the two latent codes before decoding, our RV QV can create a smooth transition from one motion to another.
>> Holy RQV.
Look at this.
March chicken lunge.
Okay. So, now is she going to do them all?
Come on. I thought she was going I thought they were going to take like multiples and add them together.
Motion bending.
>> We can interpolate latent codes from different clips to create motion clips and trajectories not present in the data set.
It's pretty crazy, right, John?
Wow.
So, I don't really know if this one has like a total theme park application, but it is, I think, interesting to see like with this technology how much faster and more affordably they can do animation, which just means for, you know, things like Disney Plus or animation or, you know, animated series, it just makes it a lot more uh Oh, you're welcome. Thank you so much for watching. Uh, I think it just makes it more applicable. So, the first one definitely had something that we could see that would work for the theme parks. This one maybe not so much.
Maybe. I don't know.
But I always watch these videos cuz like a lot of times I see stuff in there that I'm like, "Oh, that's cool." And, you know, I'm working on a video series right now where I kind of talk about how like different pieces of different attractions build the next attraction you go to. So, it's like all these technologies are like, you know, isolated. They're okay. Um, but it's when they start stacking the technologies is when you get the real real magic. All right. With the time we have left.
What if it could take you and animate you? I mean, it could be interesting though, John. Right. Like, there is a theme park application for if you were the blue person and then they could run the pink on you. Like, I could see it for like a show almost like if they had a show and they did like a volunteer like, "Hey, who wants to be a part of the show?" Oh, and you're just standing there and then on the screen, you know, you're doing uh Renmo style.
It could be cool.
Lizzy, is this the new fieldy test? When they pull you over for a DUI, you just go, "Uh, hold on. Let me just render real quick and then I can walk the line." No problem. Park mine green.
Physical animatronics. Let's talk about that after I put you in my precious infinity stones collection. All right, green.
You think you can get away from me because you're green? Look at that. Now you're red. Now you're the red stone.
Isn't red stone like a bad dude?
Oh, look at him.
Look at those beautiful stones.
Uh, so do you think this technology could be translated physical animatronics? I mean, that's interesting. Okay, so let's think about this for a second.
The first one we did where they were doing the face mapping.
I guess they could like I was thinking about it as more as like a projection anim animation, but yeah, how terrifying would it be if you were in an attraction and then the animatronic was literally doing what you were doing back at it? I don't know what the the storytelling application would be, but yeah, I think that would be possible. Park mind, I love that you do have a park inside your brain. Your avatar is true.
John Shu uh Shoemaker says, "Almost thinking it could have been for the play pavilion." Man, the play pavilion.
I knew a dude that had artwork that was going to be in the play pavilion. He was like this close. This close to having his art inside of a Disney attraction and he got taken away.
Oh yeah. All in all, just another stone in the wall.
Oh, I see what you did there. I see what you did there.
Yeah, look at you, baby.
All in all, just another infinity stone on the bricky wall. Put you over here.
No, put you over here. We're getting too getting too yellow heavy over there.
Yeah, big Thanos energy. Big Thanos energy right now.
This could be used to replace a character meet and greet. Yeah. I mean, I think with what they do with Crush and how successful that's been that there's definitely ways where this could be used for live entertainment stuff. No doubt about it.
Jay Harris, Walt Eertronic was just prepping us.
Yeah, John, I can't say, but a friend of mine had artwork that was green lit to go in there. Branos or Thicky? I think I like Branos or Thicky. Thicky makes me sound like a husky boy. I think I go with Branos.
But I like the I like being the kingpin.
I like knowing that I have one pin that rules them all.
Michael, over under 20% that tomorrow actually goes through with Avatar. Did I put up a video yesterday?
I've been seeing all the Zootopia stuff going around and I'm I think Zootopia would be a good fit, but I'm telling you guys, everything I'm hearing from the inside is its avatar. And I made the video and some people were like, "Oh, you know, like we'll see what FreshBake Dave has to say about this." I'm like, I appreciate Dave's enthusiasm, but I I mean, I have yet to have anybody from the corporation tell me that Zootopia is going in there. I just think that's fanfiction, and it's great fanfiction. I think it's what most fans would want.
They should make it a fan reality. Um, but I mean, from what I'm hearing, they are like deep into the Avatar, and I just don't think they're going to toss it out.
thicky.
So, yeah. I mean, we'll see. I I think I think that the the Avatar thing will be finally uh laid to rest based on what happens at D23.
Yeah, Jay Harrison. Like the thing is is like from what I've heard from multiple different people that don't work together is that they're working on it from the inside out. like the content aspect of it as well as the the physical space. So, um who knows could change their mind, but uh I know all all these folks are are trustworthy folks and um you know I mean nobody's told me anything major. I don't know anything major other than it's happening because I've been like that's not happening, right? People like no it's happening.
Oh, West Texan or Weston neither please.
Uh, John says, "I don't think Josh will waste all the money invested already, especially if there were contracts with Cameron." That's a good point. And they are already doing the ride footage.
Yeah, we'll we'll see what happens, you know? We'll we'll we'll see what happens. But I don't know, man. I think um there was a moment when I was bummed out about it and then when I kind of found out that it's as of right now it's the plan. Then I just kind of got excited like okay well let's just see what they can do.
Like let's like lean into it. We already know how beautiful it is out in Animal Kingdom. I mean if something like that could show up here it'll be a big win.
You know I ride attractions all the time that I've never seen the movies. I don't care to see the movies and it and in in fact I think that it you know some people get upset that I don't go watch the movies but like you got to think I have a very different relationship with a lot of these attractions because I only know these characters through uh the the rides like I only know I don't know name it like Ratatouille I only know what I know from the ride you know Little Mermaid only know what I know from the So, um, it gives you a very different experience. Like, sure, you guys get all the little Easter eggs, but like, you know, a a ride should be able to to tell the story without you seeing the movie because not everybody that goes through there's seen the movie.
Jay Harris says, "I mean, studios have dumped movies right before release, but D23 will seal the deal." I I I kind of feel like um I know that movies have been dropped, but dropping a theme park attraction, a multi-level theme park attraction, I don't know, man. It seems like dropping that for some reason cuz like you could drop a mo like movies that get dropped or if if a movie costs like over like $30 million, they don't drop it unless there's like you know horrible allegations from somebody in the movie or like somebody does like something just like horrific right before it goes out and they're like, "Okay, it's just not worth it. Cut bait." But a an avatar, I mean, granted, they haven't built it yet, but I mean, this is going to end up being like hundreds of millions of dollars, and I don't think you pull the plug on that.
And I think if they were going to pull the plug on it, you know, I think maybe that time has passed, but we'll see.
Take the animals of Avatar. Then you can have Zutar. See Sean from Phoenix. This is this is a guy who's thinking, right?
This is a guy. All right, I'll get you.
I'll get you. I'll get you.
Oh, yeah. Sean from Phoenix.
Welcome.
Welcome to being one of my infinity stones. Oh, it's so symmetric now. Looks so good.
You guys like what I've done to the place?
Well, then follow me along to Zutar.
Little Mermaid is one of the best films ever made, so the attraction doesn't do it justice at all. I don't know. I like that ride. I like it better now that it's dark.
Well, you know what, anime flicks, you're right. They did pull the plug on the Star Cruiser after a year. So, that is a very, very good point. I mean, they built that thing, but it also wasn't it was a I think that was a easy thing. I think pulling the plug on the Star Wars hotel would be easier than pulling the plug on an attraction because the hotel was an upsell. Um, and they had to get enough people to keep buying hotel rooms to make it worth it. But a theme park attraction, it's like people have paid money. They're already in the park. People are going to ride it whether they like the IP or not. And you know, a lot of some attractions don't have IP. You know, I'm sure there's plenty of rides that people ride that they don't care about the movie or have never seen the movie, but that is a really good example that they took that thing out. Avattopia, I like that one.
Steve or Stu, John is right though. They use the hotel as a massive, massive tax write off and so they can't touch it for a couple years, but then they can bring it back online for something else. and they basically get something for free cuz they already spent the money to build it. Then they took it off of their their taxes and then it's like, "Oh, we're just going to use this for offices now."
So, who knows? We'll see. Anything could happen. That's the fun of all of this.
That's the fun of everything.
Ah, the brick king pin. I love that. I love the collaboration we have going here. This is great. Oh, long live Drape. RIP Drape. Thank you, Holiday Garden. Thank you, Holiday Garden, for being real. for knowing that Drape's the hero that we all needed.
Anime flicks, they're just better locations for Avatar. I know. I know.
The video that I put out yesterday, it was funny how people were like negotiating with me like, "Hear me out, Bricky. Maybe they put it over in the Simba parking lot." I'm like, "Dude, I I I think that it's the worst place it could go." And I think it's a property that people aren't excited about. When I tell you what I know, I'm not making the decisions. I'm not saying that it's right. I'm just delivering the news to you. And so many people are like, "I don't know, man. It should be this or I don't like this." Like, I I'm just I'm just telling you what I've been told.
All right.
James Brown got him. Nobody chatted.
Nobody made that move away.
James Brown twice. Twice. Two stones.
James Brown with two stones to roll them all.
How's it look?
How's it look?
Oh, I'm gonna have so much fun with this little contraption that I just designed.
Uh, great content this week. Thank you, dude. I'm having, you know, I started off this year not knowing that I was going to be doing YouTube full-time. I knew it was not too far off in the distance. I mean, I I've done it full-time before. Um, but with the way that the economy was and things, just, you know, it was time to like step away from Lincoln and do this full-time.
And so, it took me a little bit to kind of like get the game plan kind of figured out. And then there was the force break, obviously, when I uh lost my father-in-law and was up in New York.
But since I've gotten back, I feel like the vlog has really hit a stride. Um, and I've been really really excited to see all the comments and all the like the community videos I've been doing where I just bring up kind of a topic and kind of give you both sides of the story and then you know I don't take a hard take. I love seeing what everybody has to say. Like it's been not only a lot of fun making all the videos, but it's been a lot of fun reading the comments and all that stuff. So I I appreciate all you folks for um giving me 15 minutes of your day and even more so the people that become a channel member or people that become infinity stones. Uh, I I really do appreciate it.
It takes a village and we're doing it.
So, thank you so much. And it's been really really fun to see things grow.
Um, and to find the stride. Little Adam's been doing really awesome with his two videos a week, which gives me a little bit of breathing room in space.
And, you know, like the pin video where um, he and I up the production and put sound effects and graphics into a video.
You typically don't see that in a theme park vlog. So, I'm just really, really proud of where everything's going. And I think for the last 28 days, the vlogs got over a half a million views, which is awesome. So awesome. So, thank you so much. I appreciate that. That little uh cong congratulations there, James. But of course, I have to thank you guys because without you.
Oh, look at this. Look at this. Jar Jar wants that. He was like, "Yeah, man. I want that head. I want that headlining space." Uh oh.
Nope. That's James Brown. Save the clipboard.
Yeah, baby.
Look at that stone up there.
It doesn't take much to make me happy.
Uh, West X and Bricky, you're amazing storyteller. Your narration is amazing.
You, man, you have me tearing up sometimes. You know, it's not my goal to make people cry or to make people sad like the Tomorrowland video that I made last Saturday, but it was my goal when I do those documentaries. There's there's a lot of really good content on YouTube and there's a lot of videos that are similar like the history of this or the history of that or you know like different stories you've heard time and time again. And when I do one of the documentary videos, it is my goal to tell you a story that you haven't heard before. And that's hard because there's a lot of people telling stories in the Disney space. But I kept digging and digging and digging on that one until it kind of was able to put it all together.
So, it's not my intent to make people cry or make people sad, but it is my intent to try to tell you stories you've never heard before.
Uh, I think it's a little bit too soon, Parkmind, to talk about working with Disney. Um, we definitely do that at a later point. There's still some work that I worked on that's out there that could be showing up to a galaxy not too far away. So, let's let's let's let that let's let a little bit more time happen and then one day we'll we'll rip on it because I don't want to uh I'm not worried about burning bridges for me.
I'm just worried I don't want to make anything difficult for my homies over at Lincoln.
Got to get my stones. You know, you just got, you know, a lot of people a lot of people collect the stones, but they don't know how to organize the stones, right? Got to just kind of let's just do this. We can do this now, right? We can do this.
Love it. Love it. Love the stones.
All right, let's do a couple more and then we'll wrap it up here. Uh, Trail Debunker. Hey, Bricky. Just wanted to say thanks for all your great content. I suffer from a lot of anxiety and find your videos sometimes and find your videos something I can watch, learn from, and relate to. Thank you for all you do. Appreciate you, man. Anxiety is no joke. Hang in there.
I was going to say don't think about it, but as soon as you tell somebody don't think about it, then the anxiety comes.
Oh, Alex Perez, you handsome devil, you.
No wonder you have a wife. Look at that avatar.
Oh, Alex, I know you have a wife, but she's going to miss you because now you're preserved as one of my stones.
Uh, Bricky, you have my wife and I excited for Westcot. It may be just a pipe dream, but I would love some version of Epcot here in CA. All right, let's talk about that real quick because I really think that the concept art that they showed us, obviously that's not what they're building, but I think there's some validity to that concept artwork for the Toy Story parking lot because they had to show like think about it. They just put that out and like the OC register like, "Hey, this is where we're building." But, you know, they set the mayor down. You know, they set other officials down and said, "Okay, this is what we're thinking." you know, they had to give them like the for real pitch and be like, "This is what we're really thinking." So, just imagine that this is blah blah blah blah. We'll be able to do art festivals. We'll be able to do like, you know, food and wine festivals. And because it's more of like a hangout, it's just like basically creating a shopping district, but it'll be inside of a Disney park ecosystem.
The rides will be encased in buildings.
It won't be loud. It'll generate x amount of money. Like, you know that they basically pitched them like, "Hey, this is how much money the backside of Epcot makes. This is how much money we think this will make. this is the tax base revenue, this is how we think it will work for people coming in from out of town, blah blah blah blah. Like, you know, they gave them that pitch. And that's why that concept artwork looks the way that it does. Cuz for us, it's just like, nah, it's just going to be a shopping district at a mall.
Who's building a mall right now? Um, but I think there was way way more to Why not three?
Hell yeah.
There we go.
So I think that it was basically design ridiculous.
So, I think it's basically designed as, you know, for the layman. Oh, yeah.
We're going to build them all. But for people who are on the inside, this is what we're really doing. This kind of gives you a vision into what we're talking about or what we're thinking.
Jar Jar epic. Totally epic to be three stones, dude. And so I think where would maybe be different than Epcot though if we're just thinking out loud is, you know, I've been uh doing a lot of research on Marty Scolar and they had a really really hard time um booking the uh the countries to come on board with the the Epcot pavilion. So I think that where traditionally it was like we're going to get countries to come on board, I think the new idea would be that they would make Disney districts and then they would populate it with IP that fits inside of that environment. So, um, you know, if they did something that felt like Africa, then, you know, maybe it would be, uh, The Lion King, but also like, you know, a Jungle Book type vibe, you know, somewhere in there as well. Um, you know, if they did like an Arendelle, then maybe there, you know, that would transition into other things and sort of a Nordic type vibe, although I don't know if there's anything other than that.
So, I don't think that it would be like, you know, I think that's kind of a sweet 1980s thought of like, oh yeah, we'll get countries to pay for it. But I think that if they did do it in districts, the districts would be defined by, you know, culture, obviously culture of these areas, but also like what kind of fits into the IP needs of the park. That would be my guess. That would be the logical way to do it.
Dude, that anime I've been saying this the whole time and I'll say this is the last time I've ever seen this. No, it's not. I'll say it again.
Why would Disney build a mall right next to somebody else's mall? Because all that's going to do is make the Garden Walk actually work. And what do we know about Disney? They're not in it to make it easy for anybody else. So, why would they put a mall right next to the mall? Why would they put a mall blocks from where OC Vibe is going to have all kinds of like mall type environments? Why would you build a mall when retail is dying? And people like, well, they're going to remove Downtown Disney and then that's going to become more the parks. Dude, those hotels need those amenities.
Downtown Disney isn't for the park guest. It's for the hotel guest. So on your check-in day, the day that you don't have a theme park ticket, you keep spending money inside of Disney.
So if they build a mall over there, that is the dumbest use of their last piece of land. And it is the thing that Orange County definitely doesn't need. Excuse me, I have a stone coming through.
The chat is killing the screen. The chat. Okay.
Every time I go to get a screen capture, somebody chats.
Okay.
Yes.
But I could probably do this. Let's see.
Can I do this?
with this new technology that I made.
Give me one second, guys. Bear with me.
Okay.
Oh, we are going to have a lot of fun. Who you going to have a lot of fun with my stones?
I need that image capture thing where it will keep the James Brown stone worth wherever I go.
Oh my god.
Oh my Oh my goodness.
Mark.
Oh god. Oh god. Oh god. Oh god.
Oh god.
uh Marqu says, "Thank you from all of us for being authentically real and not giving us the fluff of the parks. You taught me so much about Disney. I love your insights and look for fun details."
Hey man, I take this serious, dude. I I really love what I do and I love the people that I do for it. So, I really appreciate you saying Kevin.
Oh my god.
Hold on. Wasn't that screen capturing?
Oh, hold on. Hold on.
Oh.
Fine. Peer pressure. No mouth cut out.
Mouth cut out and beard, please.
This is amazing.
Eric, appreciate it, bud.
Hold on, let me check in.
So, anyways, as I was saying, I want to thank Eric.
There we go. Thank you, Eric, and thank you, Kevin. Appreciate it, guys, for supporting the content. You know, it's just like try to be real serious about what I do.
You know, I don't take this job lightly.
Like a lot of people like, "Oh, it's probably just fun being a theme park vlogger." But no, man. This is serious work. You know, this is really, really serious work. And if you don't take it serious, then your audience won't take it serious. And you definitely want people to take you serious. You don't want to be like a like a like a weirdo guy like, "Oh, I don't, you know, I just show up and do whatever I want to do."
No, man. You got this is you got to have your head in this game. You know what do we do tonight? We looked at all this cutting edge technology because this is a place of learning, right? This is a place of seriousness. If we wanted to joke around and be funny, guys, we go do something else. But we're all here to learn. Jeez.
So many stones, so little time. Magic space.
Oh, green one.
You think I'm afraid of the green one?
You think I'm afraid of a green one?
Look at you now. You're red.
John's like he needs a bigger screen.
All right, let me see here. Um, yeah, John, this is super serious content.
All right, I think this is where we I think this is where we say goodbye this evening. I want to thank all of my Infinity Stones. I also want to thank all my Magic Brickie holders for supporting the content. Uh, did you guys enjoy the video that I put up the other morning about how I was horny for Disneyland?
Oh my gosh. Um, I loved reading the comments on that video. So, thank you so much for uh getting the joke. And um, yeah, this is the face of someone who's serious. Culio, come on now.
I will be live again this Friday.
So, if you're around on Friday, I'll be live on Friday and I'm going to put up a poll for members and let members pick what time I go live on Friday afternoon.
So, I joined this at the wrong time.
What is going on?
I always love when somebody shows up mid chaos and they're just like, "What is happening?" Um, so, um, Cat, I watched two videos tonight breaking down breaking new Disney technology, but I created a little technology of my own where I can now place images all over the screen and not doing it through a third party, but just doing it on my own on a very, very remedial way. But this will give me the availability to have fun and and place items uh on the screen and you know we'll we'll see where we go.
Bane voice. I was born in the parks.
All right. So anyways, thank you so much. I appreciate all of your support.
Hopefully you had fun watching the videos. This was a lot of fun. I enjoy hanging out with you guys. I I always appreciate when you show up when we do really serious stuff, right? Sorcerers Workshop. really serious stuff. Looks like an old school computer attack.
Charlie Ward. Not sure if I joined late or just in time.
All right, friends. I appreciate you guys. And uh if you're a member, look for a um a poll to pick what time we go up on Friday. And if you're at the parks tomorrow, I'll be there tomorrow filming some videos before I got family coming in town for a long weekend. So, uh, hope to see you if you're around. And by the way, up to 30 pens now.
The infinity stones and the kingpin just keeps growing. All right. Appreciate you all. Have a great night. Here's Now the chat's going. When the chat goes, I can't get the screenshot.
It's stew in the How dare you get you stew? I'll get you.
I'll snipe you if it's the last thing I do.
Where is it? Oh, there he is.
I got you. I got you in my collection of stones. My inner geek is very happy.
Thanks for the geeky stuff. Yeah, man.
We'll we'll do these videos from time to time. I I think that they're kind of fun to learn from and if anything else, it just creates like something random to get into and and just kind of look at all this from a different angle. All right, everybody.
Hold on.
Have a good night.
Related Videos
OpenHuman VS Hermes AI: Who Wins?
JulianGoldieSEO
285 views•2026-05-29
BREAKING: Microsoft’s New Image Generating Model Beat Out GPT 1.5 and Nano Banana 2
aimmediahouse
122 views•2026-06-03
Long-Running Agents — Build an Agent That Never Forgets with Google ADK
suryakunju
142 views•2026-05-30
This computer is made from real human brain cells. And you can buy it.
Talktmsmedia
3K views•2026-05-28
I Made the Same Anime Fight Scene in Every AI Video Generator
NobleGooseAnime
295 views•2026-05-30
Nvidia Bets Big On AI PCs | New Chip To Power Windows Laptops | Technology | AI Updates | N18S
cnnnews18
3K views•2026-06-01
I Tested NEW Opus 4.8 on Four Projects (Updated LLM Leaderboard)
AICodingDaily
298 views•2026-05-29
3D Platformer Update - NO CAPES
SolarLune
294 views•2026-05-30











