Image-Blaster, Netflix INK & Google I/O Guesses

The animation to program the robot, they actually mocapped a real scooter puppet that was being puppeteered by a human. Mocapped him, and then put that mocap onto, put that data onto the animatronic puppet. That's amazing. [explosion] All right.

Welcome back to Denoised. Addy, good to see you. [laughs] Hey, Joey. What's up, man?

How's it going? It's been busy, I think, for both of us. [laughs] It's- It, it's been a couple weeks, yeah. Yeah.

We've been busy. It's been a busy, yeah, luckily. And also everyone I keep talking to has been pretty busy, so I'm glad, I am glad to hear that. All right.

Let's see. Let's talk about Google. I'll be at Google I/O for the first time next week, uh, so I'm gonna check that out. Awesome.

I'm excited to check that out. Yeah, if anyone's gonna be there, let me know. Hit me up. Uh, we can meet up.

But, um, yeah, this is sort of their developer conference where they announce a lot of stuff. So at the minimum, I'm expecting Veo 4 and Nana Banana Pro 2. That feels like baseline. [laughs] That's the minimum expectation?

That feels like- [laughs] Okay. That feels like, I mean, there hasn't been a big Veo update in a while. We've got Cdance 2. Yeah.

We've got Cling 3. Obviously they've been working on something. I, I saw some teases online of potential stuff. I didn't share it 'cause I, I saw one clip and I don't really know the validity of it, and I haven't seen anything else.

Can I share some rumors? Sure. Go for it. Google rumors.

So I heard that, uh, it is very close to ha- um, Veo is very close to having a high dynamic range or some crazy resolution upgrade to the video model. Okay. That is welcome. So what, some potential 4K 16 b- maybe not 16 bit, but- Maybe not.

I, I would say maybe 10 mid 4K- Yeah... is closer territory. Yeah. Let's hopefully you get to see that live.

Hopefully. Yeah. Looking forward to, uh, whatever new stuff in the creative pipeline, but then also whatever else they got cooking up. Like this video just went around from Google DeepMind.

We are reimagining a 50-year-old interface, the mouse pointer, with AI. So this demo [laughs] is- Whoa. Mouse pointer's fine. [laughs] I, this, kind of, this, you know, still makes me think of just the rethinking all of our interfaces and websites and all that stuff for- Yeah...

whatever this- Right... agentic future is. This was not fully like an agentic thing. This was more like using your mouse pointer and voice as a two-pair combo to be like pointing at something and then being like, "Hey, do this thing," and basically Gemini sort of understanding the entire context of what's happening on your computer and then taking action.

Wow, that's cool. So it's like, uh, just kinda pointed in the general direction, and then Gemini has enough context to do the action for you. One of the demos was pointing at a to-do list and then saying like, "Add emojis to that." That's not the most useful example.

But the other one was like pointing at an address and then being like, "Hey, show me the directions of like how I would get here." And then it like loads up maps and your location and then direction to there. These are very simple things, but the idea is like it is understanding what's happening on the computer and taking action. My mind is going to, um, things that I've hit roadblocks with trying to do with Claude Code, like build out a fusion node in DaVinci Resolve.

[laughs] Cl- uh, the Claude can kind of, kind of control DaVinci Resolve, but it can't build out like complicated nodes and stuff like that. Uh, or it also just- Yeah... doesn't know what it's doing. It'll like control gaps and- Yeah.

It's flying blind. Yeah. It'll be like, "I think I did it, but like I can't verify. I can't see.

I don't know." So y- Exactly... you have to check. So if this is a step towards that direction too, where you can kind of have better visual understanding of what's...

Or, or the AI can have better understanding of what's happening on the computer to help you do things. Interesting. Yeah. I think the important here is the, is the interactive feedback type so the AI system can see and kind of understand and read what you're trying to do, which is important to have a, a two-way communication between the AI system and the computer itself.

Yeah. And in- interesting that it's... 'Cause I, you know, we've heard a lot about just text control and speech-to-text or just voice control, and this is not totally throwing out the mouse pointer. It's just like what if you use pointing and voice in conjunction.

Yeah. I would love for this to be in Blender or some crazy software where, you know, you don't have to really know it. You kinda just point the mouse- Yeah. And-...

and just kinda say. Y- yes. We've, and we've seen f- f- f- like in our last episode where we talked about Claude just taking over everything [laughs] and integrating- Yeah, yeah... with everything.

But yeah, there is a limit to how far you can get just through, uh, text prompting Claude and telling it what you want. You need that back and forth dynamic, and this feels like it gets a little bit closer to that back and forth dynamic. Totally. Okay.

All right. I, I take it back. I was m- I was like, "Hey, why mess with the mouse?" Okay.

[laughs] All right. I guess... [laughs] Other thing from DeepMind, it's still in a research preview. It is called Fabula, and it is a story writing AI assistant that they- Uh-oh.

Uh-oh... worked with and developed it with a bunch of playwrights and screenwriters and based it- Oh... off of story structure books. And, um, the idea is for it to be a sidekick as you write, build out your scaffolding for your story structure, and sort of this, again, not the AI doing the stuff for you, but the AI assisting you.

Yeah. The, it's like a assistive agentic story writer that helps you- Frame. Yeah... build, build a hierarchy of the story, but not necessarily come up with ideas and things like that.

Yeah. Or helping you refine beats or story beats or building out character structure and then that, that, that scaffolding. Uh, so as you do flesh out the story, you have this baseline that you're working with. Okay.

Two things. I, I love this. I, I love that they're thinking about this. Of course, this is a precursor to, let's say, a video model, right?

Like, this could potentially be an end-to-end pipeline that's fully assisted, um, uh- Story... agentic. Storyboard. Yeah.

Shots. Storyboard. Video. Shots.

Yeah. Cutting, VFX, blah, blah, blah, blah. Now, uh, two things. One, I'm gonna cite, uh, Ben Affleck.

He was right about this. [laughs] I think he was on the Joe Rogan show or something, where he was like, "Look, all these agentic story building things are just gonna deduce down to the most average and the most overused story formats." Um, so how do you create a story that is asymmetrical, that is completely unconventional? Like, how you create Memento?

That's backwards. Yeah. Yeah. Right?

So, like, that's where the art lies, is the outlier stuff. And so, yeah, I think this stuff is probably not gonna be very useful for somebody who's trying to really break the mold, but if you're generating 100 videos a week, you're a YouTube content creator [laughs] you're doing verticals- You're gonna be using the same [laughs] Yeah. [laughs] Verticals are gonna kill you. [laughs] Let's go.

[laughs] What? Let's go. [laughs] Give me the slot machine. [laughs] Christmas in random town- Not even a slot machine...

in a small town. Just generate all 100 of them. [laughs] Yeah. I don't care.

Just put it out there. Yeah, you know, you just reminded me of something. Before the AI days and all this stuff, you know, there have been tons of screen- I, I can't even remember those days. [laughs] From a time long ago.

There were tons of, you know, screenwriting books and, like, this is how you write your screenplay, and, like, this is what you have to hit in the first, this is what you have to have on page one, this is what you have to have on, like, first act, and this is what you have to have, like, second act. Then there was always a chorus of, you know, screenwriters of just, like, yeah, if you want to write something formulaic, then, yeah, follow those screenwriting books, but if you want to be original- Yeah...

then, like, throw those books out and, like, write something completely different. Yes. And this is just the AI op- [laughs] This is the AI version of the screenwriting book, so it's like, cool, the, it helps you if you want the help. Yeah.

And if you don't want it, then yeah, sure, it will give you something that follows a, a formula pattern. You could argue, like, um, you know, I mean, that was here with 1000 Faces, where, like, as humans, we have the same story, uh, archetypes that we've been doing for thousands of years. Regardless, if you want to do something completely different and throw it out the window, like a Pulp Fiction or Memento, um- Exactly. Exactly...

you know, that's where humans, that's where- Yeah... you know- Okay. I, I have a-... human creativity comes in...

I have an idea on how to be super creative for really- [laughs]... really cheap and very easily. [laughs] Okay. I call it the Jim Cramer Fabula.

[laughs] Do you get it? So you have Fabula make the story and just- [laughs]... reverse everything. [laughs] Like the anti ship.

And now you have something amazing. [laughs] I think you could do that today too, if you just go to ChatGPT and you're like, "Give me 10 ideas," and it gives it to you and you're like, "Thank you. Garbage." Garbage.

Yeah. Like the opposite of those 10 ideas- That's the do not write list... is my idea. [laughs] That's it.

That's, that's how you become famous. All right. [laughs] Jim Cramer. The Jim Cramer- Jim Cramer Fabula...

School of Screenwriting. [laughs] Okay, good, I like that. Okay, this one, this one I've been texting out to people [laughs] and everyone's going crazy over it. Yeah, dude.

Uh- I, I nearly, uh, blew my mind over there. So we have talked about World Labs, we've talked about generative text to 3D. Nielsen, uh, who I believe works at World Labs, yes, built an open source tool that is basically a, open sourcing a 3D gen toolkit for Cloud Code. You give it an input image, and then it is basically pulling World Labs' text to 3D models, which we'll get into in a second, and using Cloud Code to build out the full 3D space, but with your 3D models, with the world, with lighting, physics, audio.

[laughs] So crazy. This is, so this is like World Labs on s- on steroids. Uh- Yeah. [laughs] This is, this is somebody at World Labs who used Cloud Code to its full potential.

And open sourced it. Thank you. [laughs] Yeah. [laughs] Pretty...

There's, you can get the, you can download the... Again, I have not messed with this yet, but this is on my, high on my list, because this could potentially solve- This is-... a lot of things I've wanted to try to do... buddy, this is exactly where AI should be heading.

The skills harness a- against an input image and generate meshes with a combination of image generation and 3D generation. Sure. I'm assuming you could also, I think it's calling up different file APIs, so I'm assuming you could also plug and play your favorite text to 3D mesh model of choice. Mm-hmm.

Mm-hmm. That you might not be locked into anything. World Apps is kind of the only thing that will do, like, a full 360 environment build. Yeah, but, but like for them to decompose a world into individual 3D models, that's pretty freaking hard.

And finally, there's a companion app to view and pose all the objects that were created. Oh, wait, there's like an actual app that you can do this into? What? Yeah, and it has physics.

Like, do you see- Yeah... that thing get tossed around and bounce off the floor? Yeah, it's throwing around a chair and stuff, like as if- Hell... it already has all this stuff built in of what it should be doing.

Um- And I think also audio, right? If I'm not mistaken. He said audio. I don't know where the audio- Like if the-...

comes in... the chair hits the floor, there's a sound generated. I guess. I mean, you're getting really close to game engine territory here.

So it's called Image Blaster. Uh, creates 3D environments, sound effects, and meshes from a single image using Claude Skills, World Labs, and File. So yeah, I mean, this is like basically built this, your 3D world, but now it has models, it has texture, so you can move things around, you can edit it, you can modify it, but you're just getting that, like, crazy fast, good starting point. Yeah, I mean, call me old-fashioned, but this feels perfect for Previs.

Build this world- Yeah... very, very quickly, put the, put the world up, and then start moving things around, put the camera where you want it, and then that goes into a image model or a video model for traditional 2D output. I mean, you, definitely Previs, or definitely, you know, I mean, I'm still on this path of, like, can we... You know, this is like a, a next level up of gray boxing.

Can we kind of quickly iterate a world and then feed that into some gen AI pass, and have the consistency of a 3D world from different angles, but then we get to final pixel faster? Exactly. [laughs] And I, I think in 2026, we're mid 2026, that feels like everybody's on the same page with that is, like, text to video or text to image is not quite the control that I need. It's not working.

Mm-hmm. I do want the traditional 3D controls, but I don't want to mess with rigging, I don't want to mess with key frames and animation. Mm-hmm. Um, so give me something easier there.

But then I love the final quality that generative AI gets us, so finish it- Yeah... with that.Yeah, finish it with that, but at least I have my consistent world, so when I give you these inputs, I am getting that consistency, hopefully, in theory. [laughs] That's...

Still got some kinks to work out on that one. Yeah. But that's the idea, or I think where this is going. Yeah, so wild tool, open source.

I will put the link in the description, but it's called Image Blaster. Shout out to Nelson. All right, now let's talk about some of our favorite things. Uh, AI pitchforks.

[laughs] Someone- Oh, that's a Van Gogh. No, Monet or Manet. Monet. Someone- Monet.

Okay... yeah, Schlomes posted this image where he said, "I just generated an image in the style of a Monet painting using AI. Please describe in as much detail as possible what makes this inferior to a real Monet painting." Oh, you're just asking for it.

And so then, you know, unsurprisingly got a bunch of critiques of saying, um, you know, how it's terrible, how it's wrong, how it's crappy. Yeah, you trained on Monet data. Of course it's gonna be exactly [laughs] like it. Turns out, um, it actually is a real Monet painting, and he just said it was AI to trigger people.

And [laughs] That makes- [laughs] This is just a style transfer... and so [laughs] so this traction research of people systematically downgrade their aesthetic assessments of art when told it's AI generated. So basically, I th- we have talked about this before. As soon as anyone hears that AI is involved in something, they get upset regardless of- Oh, okay.

So that was spoofed in order to get actual, uh, scientific data on how people take things- Yeah, he posted this image saying it was an AI generated in the style of Monet, but it's real Monet. It was a crop, this is a cropped section of an actual Monet painting. This is Monet. That's what I thought.

Yeah. This is not AI generated. Okay. But then all of the comments were saying how there's a certain harshness, no soft blending of colors, no depth, no s- symbiosis of the elements.

It's all Borked nonsense with no sense of space. [laughs] No frame, no sense of the threshold between subject and object, just colors. Okay, no frame because it is cropped. I give them that.

Okay, right. It is cropped, so sure. Yeah. It is cropped.

There is that crop. Maybe that affects the sense of the depth. But um- Colors are more pleasing to the eye. The brush strokes are more delicate.

Uh, del- The water looks more tr- oh, okay. Good, good notes. All right. You can tell the difference.

You can tell the difference. It's extra different flowers. [laughs] Extra different flowers. [laughs] I love it.

As, uh, y- yeah, so basically posted it, said it was AI, people got upset, said it was crap. Yeah. Turns out it really is a real Monet, and when people hear things are AI, regardless of the quality of the output, um, people get pissed about that. Wow, that's such an interesting study, man.

[laughs] Um, uh, I'm really glad you brought this up. I mean, that, that kind of applies to our industry in a lot of ways. I mean- We've encountered- I know people on both sides of this fence. I mean, you know, this is a...

[laughs] From like when we first kicked off the podcast when, um, um, the, uh, the brutalist crap was happening. Exactly. You know, and it was like for the, the cl- oh, clean up the, a little bit of the dialogue. Oh, used AI.

Oh, whatever. That upset people. And then, um, yeah, all of the other stuff. And, and I think we talked about too how like Netflix has been playing it smart when they're like after the fact mention, "Oh, by the way, [laughs] we used AI for the VFX on this," months after, after it aired.

Right. So that people weren't like, "Oh, like I'm gonna- Yeah... not watch this thing." And, and, and they're- It's like-...

basically preparing and softening their landing, right? For the time when the eventual announcement comes that this movie is half generated in AI. I'm sure that's gonna happen sooner or later. Yeah.

Well, we also got that story coming up in a second of the animation studio. Um- That's right. But yeah, the, the, yeah, I mean, I feel, you know, you've probably experienced the same thing. Like this year, 2026, has definitely been a big vibe shift just from everyone in general- Oh, yeah...

of more openly talking about AI. Um, and anyone who is skeptical or dubious before has, for the most part, at least been like, "So tell me about this AI stuff." [laughs] Like realizing it's not going away, and now they kind of gotta get an understanding of it. Um- I mean, like I, I'm gonna play the other side of it for one second.

I, I do understand- Yeah... the fear of somebody who spent their entire career just being really good at this one thing, right? Like when I was at DreamWorks, we had career modelers. Like they, all they literally did was ZBrush for like 10, 15 years, and they, they're just so good at modeling 3D assets that they can charge people $50,000 for an object.

But it's the best version of that object digitally. Mm-hmm. And overnight, their career, their craft, their expertise is more or less futile. So what else are you gonna do?

I mean, even if you try to learn the new tools and sort of apply your modeling knowledge to this new world, you're still not gonna be as respected, as paid as you once were. So there- Yeah... there's, there's certain parts that I, I do get. Having said that, like not all of us are expert modelers, and like we honed in on this one craft our whole life.

Most of us, I think, are generalists, and we're pretty well adaptable. Um, so and that's you and me for sure. Uh, and I think there's a, there's a place for us in this new world. Yeah, I mean, the specialist, you know, it's like they still have the eye.

You have the taste. It's like how can you translate that over to, you know, whatever this gen AI- Yeah... spews through. And, and somebody was saying something- Looks like...

really interesting to me the other day as well. They were like, "Hey, um, when we went from 2D animation like The Lion King to the first 3D, like Toy Story, and Shrek One, and so on." Mm-hmm. Like there's a lot of 2D artists that were basically out of a job because they didn't, they didn't pick up computers.

They were just- Yeah... they did, you know- Same story with, um, uh, I'm blanking on his name. I know he's the super famous motion, uh, stop motion animator, uh, you know, with Jurassic Park. When he did a stop motion test of a, of a, of a T-Rex run or a velociraptor run, and then they did a CG test, and then he saw that, and he's like, "I'm out of a job."

He's like, "Yeah." He's like, "I'm-" Game over. He's like, "I'm, I'm..." Yeah, he's like, "I'm extinct now."

[laughs] So you know- Yeah... this, you, you was world-renowned stop motion animator, but it didn't- Yeah... the world kind of shifted, so it's the same. Yeah, there's just this constant churn- The same shift every 10, 20 years...

of people that just kind of get left behind, and then some people that get elevated at the same time, and then, and so on and so forth. But this, this churn that this time around I think is the most amplified and the most, uh- It's amplified... impact- impactful. It's also compounded, and an easy punch your bag, we talked about this before, of just like other economic issues-Completely outside of AI- 100%...

obviously just- We're just not in a good place- LA production-... in general for this. Yeah. Yeah.

Exactly. If, like, LA production was churning and was what it was- Yeah... early 2000s, like, you know, maybe this would not be such an- Five years ago, LA production was fine... issue.

But, yeah. Yeah. And streamers were throwing a bunch of money at everything. Anyways.

Yeah, okay. Monet. [laughs] Monet, yeah. Well, wait, actually- I'm, I'm a fan of- Okay, we got another good...

We got another- I'm a fan of the real Monet, not the- [laughs] We got another good AI- [laughs]... AI punching bag story. Yeah. Okay, so this post came out from Sony Xperia, which honestly, I'm like- The phone?

Oh, is that their phone brand? Yeah, yeah, they, they- Yeah, okay, their phone brand. Yeah. Those are some nice- Oh, yeah...

cameras on in, on those phones, dude. Have you ever used one? Well, I have not. Mm.

But- Chef's kiss on those cameras... they posted this yesterday saying, "The new AI camera assistant with Xperia intelligence brings stories to life." And they showed this demo of [laughs] these before and after photos of [laughs] before- Ugh... and- What?...

after with the [laughs] AI camera assistant. [laughs] If you're listening, the, uh, AI camera assistant after photos look way worse [laughs] than the original. Completely over-processed, uh, blown out. Over-processed, overexposed.

Yeah. But yeah. The original retains- Just a sandwich... so much detail, shadows, nuance.

The, uh, the AI one feels like s- AI slop. [laughs] So, um, you know, understandably, the, uh, internet had a fun-filled day with this. The Michael Ja- Uh- The Michael Jackson one. [laughs] Come on, dude.

Show the, show our viewers that one. All right, fine, I'll show it. I'll show it. [laughs] All right, I'll show this one.

[laughs] Okay. [laughs] So, okay. Um, but so they, um, f- posted a follow-up post about this today. Oh, okay.

They apologized? "Following the post about AI camera assistant, we'd like to explain the feature in more detail. It doesn't edit photos after shooting. It suggests four settings and different creative directions based on the scene and subject.

You can choose any option or use your own settings." Too late. So- You ruined the brand... that was a horrible demo.

That was an awful demo of what they did before. This makes much more sense. This is like a, this is like an auto exposure with the, you know, more scene awareness, and it's, like, trying to do some auto exposure plus whatever settings to guess how you wanna take the picture. But what's, what's the AI part in these four suggestions?

I don't see outpainting. I don't see new objects. No, I think... 'Cause it's not even...

It's, it's literally, it's saying it's before you take the picture. So, like, you know- Oh... auto exposure would, like, look at your light meter and then adjust your, like, shutter and aperture based on your settings. This is taking that to the next level where maybe it's adjusting your color settings.

Maybe it's doing something else before you hit take photo. All right. I guess, yeah. So it's like, yeah, that makes more sense.

It was an aw- that was a horrible post. [laughs] That was a horrible first post. That was a terrible post. This makes more sense.

Like, you use it, you don't use it, it's, it's like, uh, auto exposure on steroids. I, I understand that more. Right. Um, terrible launch.

A terrible launch [laughs] to try and- I would love a feature like this on, like, a professional camera, like a RED or a Sony Venice or something 'cause, like, nine- They're gonna take you out back and-... out of 10 times when-... you know, beat you up over that comment... unless you're Greg Fraser- [laughs] It's like-...

you're not exposing the camera right, and you're not putting the light in the right place, you know? So to just get suggestions on some of these things, uh, before you hit that, you know, record button and record away one terabyte, this is highly useful. I mean, I think what you're talking about is not just a auto thing on the, on the RED. [laughs] No.

No. But- No... but a unified system, a- your aperture lights and your camera and the vision- Exactly... you see is all connected- Yeah, it's just-...

and- It's a look choice... adjusted. Um, so you have four different looks you can pick from. Mm-hmm.

And each look will affect, uh, the LUT, you know, the ISO and everything, and you can kinda play around and just experiment more with, uh, principle photography. You heard it here first. Addy wants an automatic RED camera. [laughs] Yes.

Give me those auto settings. No, I want a Sony Venice 2. Well, I don't know, maybe you could, uh, maybe you could direct- If we get a Sony Venice 2- Give me an ARRI with-... then, um, we'll talk great things about the Xperia on the, on this pod...

I want an ARRI with autofocus. [laughs] What's that new camera? What's the Nikon that has the Z sensor? 'Cause that might have auto stuff in it.

The one that went to space? The one that they just laun- their first, the first new product that came out of the Nikon-RED acquisition. The Z9, I think. It was like that...

Wasn't it, like, that 2K camera, that $2,000 camera? It was like a- Oh, the little guy. Yeah, it's gotta be, like, a Z90 or 900, something like that. Yeah, this one feels like that would be your RED with some type of- Smarts...

AI assistance. Yeah. Smarts. Smart RED.

Does Sony have an AI department? I mean, I'm sure they do. Like, for what realm... They have ZIN, which we covered at NAB- That's right...

which is, like, their spatial capture ecosystem. They have... They did release a broadcast camera. I didn't talk about this, but they released a broadcast ENG style camera that had a content authenticity ch- credential built in, so, like, as you're recording with the camera, it's...

I don't know whose system they're using or if it's their own system, but it's, like, embedding data that, like, says this was, you know, legit. Like, this is a real person on their- Actually shot video. [laughs] This is, this is real, authentic video. I don't know what other software you need to, uh, verify and read that, but it was, like, the first camera that had that built in.

We've talked about this before of, like, the different content credential systems, um, that would be embedded in cameras to verify that footage was actually shot and m- not manipulated with. Right. And the same thing goes for generations. If, if it's an artificial generation, then the, the watermark should say that on the CAP, right?

It should. I mean, yeah. [laughs] Or who- I know Adobe has that a lot in stuff. Yeah.

I think the onus is gonna be more on verifying authentic material versus tracking generative material- Oh, I see... 'cause that's way easier to- For, like, news and, like-... to scrub... world events.

Yeah. For news and verification- Right... and authenticity of- Yeah... like, did some- you know, did someone actually film something happen?

Was this an actual thing? I think-The verification's gonna need to be on the verifying real material rather than verifying AI stuff, 'cause you could strip that pretty easily. Sure, I can see that. And specifically on broadcast cameras, 'cause that's what goes out into the field.

Yeah, that's exactly, that's why it was in an EMG style camera- Right... not like a point and shoot. Not, not like a cinema camera. Yeah.

Yeah. [laughs] Exactly. Our job is to make stuff up- Yeah. [laughs]...

versus broadcast is to capture the truth. Right. [laughs] Yeah. Exactly.

Exactly. Yeah. Right. All right, what else we got?

Okay, so coming back to that Netflix story. So Netflix launched a AI animation studio called Inc. Yeah, I, I heard about this, uh, while we were not shooting episodes, Joey. We could have talked about it earlier.

But yeah, this is, this is big news because this is, this is not, like, a team or a department. This feels like an entirely new studio that they're spinning up. Uh, it's interesting, too, 'cause it's, like, also, like, where's the line of, like, what Netflix is producing in-house versus where they're getting outside vendors, and this feels like a completely in-house vertical thing. Yeah.

Yeah, sure. I mean, like, uh- To start producing content... the scale at which Netflix operates, I think it's impossible to make everything in-house, right? So they'll ha- Right...

they'll handpick, like, the top 5% or 2% of the shows that they're gonna make in-house, and then the rest they absolutely have to go to vendors for. So my guess is this new studio is probably not gonna be attached to anything, anything substantia- like, anything that's gonna air on Netflix anytime soon, but they are gonna start to spin up the production methodologies and to get the quality bar up and figure out how to actually make something with AI before that goes out into the streaming service.

Yeah. Uh, a little bit more in the article. Uh, they, they haven't posted anything publicly about it. Most of this info is sourced from job listings.

And so, uh, Inc-ubator's long-term technology strategy will focus on, quote, "GenAI-enabled workflows, artist tooling, and scalable, secure, multi-show environments," according to one listing. Uh, another listing says that they're about creating animated shorts and specials using experimental GenAI native production pipelines. Mm-hmm, mm-hmm. So yeah, starting short stuff for now or experimental stuff to prob- validate the workflows and then moving to bigger things.

Yeah, I mean, if you look at the budget for Blue Eye Samurai, I think, which was, like, 100 plus million dollars, you can now make a lot of those shots in GenAI. How much savings would that be? Like, I think these are the questions that they're asking internally. Blue Eye Samurai's so good.

One of the best- I love that show... [laughs] one of the best animated movies ever made. And also in- Yeah. [laughs] A serie- a serie.

It was a mini-serie. Well, it was a season. Yeah, right. Yeah.

And then, uh, but yeah, it has taken years for season two to come out. [laughs] Well, like, the difficulty with Blue Eye Samurai is it's highly, um, adult content, but it's animated, and so, like, the audience is really small. Mm-hmm. But I've, uh, but no, I think it, yeah, it's not a cheap show, and it- Yeah...

it just also takes a long time to make- Yeah... at, with current tools But, like, that's an example of complete in-house Netflix production. Oh, was that all done in-house? Mm-hmm.

Yeah. Okay. One of the few stuff. It's interesting that they chose animation before live action to do the, um, first AI studio.

Ooh, I think that's, it- it's, we know that's easier to do now. Yeah. Like, that's pretty- That feels like a more achievable goal. Although, I will say- Yeah...

once you get into the mechanics of it, there's so much hand-driven art in animation that AI still probably doesn't have yet. Mm-hmm. I mean, maybe Netflix does, who knows. Mm-hmm.

But things like squash and stretch and, like, the timing of animated characters and, like- Yeah... the, the way their mouth exaggerates when they have an expression. Like, you know, it's not enough to, like, drive, um, [tsks] a character with, like, a video of yourself driving the character. Like, you really have to, uh, you know, go above and- Right, 'cause the animated characters move in ways that- It's, it's crazy.

Like, how expressive it has to be... physically not possible. That's not work. And it, it's beyond puppeteering.

Yeah. It's, it's an art form onto itself. So, like, having the training data for that to train a model to exaggerate like that and then to control it in a way that is meaningful for that story moment, that's the challenging part, and that's the stuff they have to solve at Inc. And yeah, I think that's, I think you're right on the money with, with building out these tools to figure out that level of control that animators need.

Mm-hmm. All right, two quick ones. Actually, this one ties into animation, and this was just kind of a cool story breakdown. Um, so at Disney World, Florida, there was the Aerosmith Rock 'n' Roller Coaster.

Uh-huh. Are you familiar with this? No. It's been Aerosmith branded since they opened it, whatever, mm, long, mm, 20, 30 years ago.

They're rebranding it to, like, a Muppets theme, and they're redoing the, like, show opener stuff. And so there is an animatronic, uh, scooter that's, is, like, you know, part of the pre-show, and so, like, it'll be a permanent robot. But the animation to program the robot, they actually mo-capped a real scooter puppet that was being puppeteered by a human. Mo-capped him and then put that mo-cap onto, put that data onto the animatronic puppet.

That's amazing. And I was like, like, I was like, "That's really super clever." [laughs] That's awesome. Yeah, 'cause I was looking at the picture of this puppet in a mo-cap suit.

I was like, "Is that just, like, a costume?" Turns out that's a real mo-cap suit. [laughs] That was... He's got a little mo-cap hat on.

Yeah. [laughs] He's got markers on his face. They're tracking, like, the mouth opening and stuff. I think that's just great 'cause that's, like, also this good hybrid of, um, you know, you're still able to keep that human touch- Yeah...

in animation that the Muppets are, are, are, are so ma- famous for, but translate that over to a programmed, uh, robot. Oh, I love this, Joey. This is heartwarming. Which they probably weren't able to do.

Uh, you know, I don't know. Uh, obviously I was, grew up in Florida, so have deep ties to Disney World growing up. And so there was the Muppets 4D show that was always great, kinda fell into disrepair, and then they shut it down. And so the, their, like, trade-off was like, "Well, we're gonna rebrand, uh, Rock 'n' Roller Coaster to The Muppets."

But The Muppet Show had permanent, like, robots that would kinda perform. And so I'm thinking like, "Oh, those robots, animatronics never had the same level of fidelity that you could do now with both the improvement of how robotics have gotten- Right... and then also this mo-cap thing. So that's cool.

That's cool. Beautiful, beautiful, beautiful execution of technology for entertainment. Yeah. And shout out to Imagineers- Yes...

for always not getting it out of the park with that stuff. [laughs] They never fail to disappoint, do they? Yeah. Yeah.

They're always, it's always great stuff. They're all wizards. Yeah. It's always fun.

Cool. All right, and that is it for today's show. Yeah. Uh, links to everything we talked about, as usual, at DenoisePodcast.

com. I have nothing but thank you. Thank you for watching and listening to us. A little over a year, we're getting close to our 100th episode.

Joey, I think that's two or three episodes away. Damn, man, I have not tracked this at all. [laughs] So [laughs] I have... All right, we'll have to look at the actual number.

Yeah. Um, yeah. We'll figure this out. This is a special one.

Damn, now there's, like, pressure. I gotta think, we gotta think of something. Oh, yeah. [laughs] Leave us a comment if you can think of something.

Why'd you bring this up? [laughs] 'Cause we, um- Yeah. [laughs] Now we gotta think of something. Now we got work.

[laughs] All right. All right, thanks everyone.