
Denoised · 2026-05-05 · 43 min
Key moments - from our scoring
Substance score
54 / 100
Five dimensions, 20 points each
Joey returns from NAB 2024 to discuss emerging tools reshaping creative workflows. Lightcraft's Spark platform emerged as a standout - a browser-based pre-visualization system that lets users Gaussian splat scan physical locations, then use iPhones or virtual cameras to move through these environments for blocking, storyboarding, and camera recording without requiring Unreal Engine setup. The conversation touches on Selects, which integrates Claude and MCP servers for AI-assisted podcast and YouTube editing with automated B-roll placement and agentic workflow potential. Martini impressed with its canvas-based (not node-based) editor supporting multi-user collaboration, timeline editing, and a proprietary angle editor that analyzes frames to create 3D spaces where users can reposition cameras and generate new angles. The hosts also cover Frame.io's announced two-tier storage pricing model and broader industry trends like cloud storage cost increases, the heating competition among cloud-based media tools (Shade, LucidLink, Sweet Studios, Frame Drive), and the physical infrastructure challenges of supporting AI workloads. The discussion reveals a pattern of companies thinking agentic - building tools that AI systems can interface with directly.
Lightcraft Spark is a browser-based virtual camera platform that lets you Gaussian splat scan a physical location, import that scan, and then use an iPhone or virtual camera to move through the space in real time for pre-visualization, storyboarding, and camera movement recording - without requiring Unreal Engine setup.
Selects is more targeted at podcast and talking-head editing with stronger Claude and MCP server integration for agentic workflows, while Eddie AI excels at creative narrative editing with voiceovers and B-roll; both automate tedious tasks but require human judgment for artistic decisions.
Martini's angle editor analyzes AI-generated video frames, reconstructs them as 3D spaces using Gaussian splat-like techniques, and lets cinematographers reposition a virtual camera to generate new angles and shot compositions from a single frame.
Frame.io is creating separate pricing tiers where the cheaper standard storage is cloud-only, while premium-priced storage includes the ability to mount files locally on your computer like a local drive.
The major competitors are Frame Drive, LucidLink, Shade, and Sweet Studios, all offering cloud storage, archival, and mounting capabilities for editing workflows, with Shade being positioned as slightly cheaper than Sweet Studios or LucidLink.
Our reviewer’s read on each dimension, with quotes from the episode.
The episode covers multiple NAB products and announcements with decent technical detail on several tools (Lightcraft Spark, Martini AI's angle editor, Selects' MCP integration), but much of the runtime is spent on casual banter, setup explanations, and tangential discussions (Vegas stories, data center tangents, Wishbone references) that dilute substantive content. A B2B operator would extract useful information about 4-5 specific tools but wade through significant padding.
So the idea is it's a web platform. You can Gaussian splat scan a location, bring in that splat into the, the platform, and then it, you can connect your iPhone or virtual camera and basically have an Unreal Engine Vcam, but all- browser based
they have this integrator that, you know, is basically like you can talk to Claude and start having it build out, like, elements and worlds and gray boxing and other things in Blender
The episode recaps announced products and industry news with straightforward summaries but limited original analysis or contrarian takes. Observations about Claude integrations, Kling's 4K support, and frame-interpolation workflows are competent but largely confirmatory rather than revealing non-obvious insights. The discussion of Blender's rejected sponsorship touches on interesting tensions but doesn't dig deep into systemic implications.
And if you go a step further, imagine tying a VCAM, like a Lightspark- Mm-hmm... a li- um, you know, uh- Sorry. Lightcraft- Blanking... Spark
I think Blender has a, has a huge place in the future of, uh, generative AI. Yeah, you need that control. You still need that, you know, at a certain level, you need that fidelity and control
This is a two-person host conversation with no guest. Both hosts are practitioners (Addy in AI/video production, Joey in editing), but the format lacks the depth that comes from bringing in a specialist guest with deep expertise in a specific domain. While the hosts have relevant experience, the episode reads more as two friends recapping a conference than a structured interview with someone who built one of these tools.
We'll talk about it on the next- Yeah. The one thing I didn't see that I saw some videos pop up was, uh, someone had built a iPhone bullet time rig
We... It was the magic of the cloud. [laughs] Got remote GPUs.
The episode includes several specific product names and technical details (Lightcraft Spark's Gaussian splat scanning, Martini's angle editor, Selects' Claude/MCP integration, Kling's 4K support), but lacks hard numbers, pricing data, timelines, or measurable outcomes. Anecdotes about Jacob's shipping-container stage and NAB attendee feedback are mentioned but not quantified. The discussion of Blender sponsorship (€240,000 for four developers) provides one useful metric but most claims remain general.
So they're supposed to launch it later this summer
they have this integrator that, you know, is basically like you can talk to Claude and start having it build out, like, elements and worlds and gray boxing
The hosts engage in friendly back-and-forth with occasional follow-up questions ("how responsive is it?", questions about Claude's visibility in Premiere), but rarely push back or challenge claims substantively. Much of the conversation is affirmative banter rather than probing dialogue. Questions exist but lack edge; the hosts accept most product explanations at face value without deeper interrogation of limitations or trade-offs.
One, how responsive is it? In the demo there, and this is also NAB crappy Wi-Fi, it was responsive
I am curious. I'll mess around with this with, like, Premiere. I'm curious if this is aware of what it's doing, of what it's able to do. Yeah, I'm guessing no
Computed from the transcript - who did the talking, and the words that came up most.
Lightcraft's browser-based virtual camera, Claude integrating with Blender and Adobe, Kling AI hitting 4K - here's what shifted in film tech while we were on the road. Joey returns from NAB with a roundup of under-the-radar tools worth paying attention to, including Lightcraft Spark, a browser-based previs platform that lets you scan locations with Gaussian splats and move a virtual camera without touching Unreal Engine. He also breaks down Selects by Cutback, an AI-assisted editing tool building toward agentic workflows via Claude and MCP servers, and Martini, an AI filmmaking platform with a standout angle-repositioning feature that turns a 2D frame into a navigable 3D space. - The views and opinions expressed in this podcast are the personal views of the hosts and do not necessarily reflect the views or positions of their respective employers or organizations. This show is independently
Transcribed and scored by The B2B Podcast Index.
I could also, like, guess that the people upset about this have never touched Blender in their life and- Yeah. [laughs] 110%. [laughs] Like, do you know how hard it is to do anything in Blender? Yeah.
[laughs] Just your AI. Yeah. They're like, "Ah." Yeah.
It's like, okay. Yeah. Like, please go use Blender. All right.
Welcome back to Denoised. Addy, uh, it's been way too long. This is all my fault. [laughs] Hello, stranger.
Where have you been? [laughs] I have been stuck in Vegas Groundhog Day. Yeah, NAB, and then a completely separate project, but I was still in Vegas for, like, a whole other week. Um, [laughs] which was way too much Vegas.
[laughs] Yeah, that's too much Ve- I'm, I'm glad you're back. You need some detox with, uh- I, I, yeah... Santa Monica juice press places. Yeah, I need some West Side fresh air.
[laughs] Yeah, and after Vegas, and then it was, uh, we just filmed the Cinema Synthetica competition this weekend, which is playing at AI On The Lot. Um, this is definitely the biggest year. It's like the 48-hour AI film competition. Definitely biggest leap in quality, which I expected, 'cause the models are just- Yeah...
so good between Kling and, um, Cdance. So. Yeah, you've been covering that conference for a while now. So you, you- Since they started, yeah, the first competition they did.
Yeah. Uh, which also is sort of how I also got into AI in the first place, 'cause I was filming them doing stuff, and then I was like, "All right, well, I'll give it a try." And then, um, fell down the rabbit hole. [laughs] That's it.
Yeah. So glad you're here on this side. [laughs] Yeah. [laughs] Yeah, so I mean, NAB, uh, we'll kind of do a little recap of some cool things I saw.
But, um, yeah, definitely saw a, a bunch of fans of the podcast, so it was great seeing everyone. There was, uh, one guy who stopped me, Jacob, is building a, or he built a virtual production stage in a shipping container, um, [laughs] in his backyard. Wow, okay. And he's like...
And it was, like, really awesome, really impressive. He showed me some photos. And then he also built, he's deep into Comfy, and built this crazy Gaussian splat Comfy workflow that kinda does World Labs, but, um, he's like, "It's better quality than World Labs." So c- really cool for showing that.
I'm curious to learn more about that workflow. If, Jacob, you're listening, you got a link, send it over. Yeah. Uh, if you wanna show that.
Wow, Jacob, you're a mad scientist. Yeah. [laughs] Shout out to you, man. Also, yeah, uh, there were a lot of people also saying, like, they don't listen to podcasts, but they listen to our podcast, [laughs] which was super nice.
And [laughs] I- Is that good or bad? [laughs] I think it's good, but now I feel a little bit more pressure, 'cause, uh, you know, it's one thing when you're just talking to the void into this camera to you, but now it's putting faces to people listening, so. [laughs] I mean, the other nice compliment was that someone said they listen to three podcasts all in: a16z- Good one. [laughs]...
and Hearts. Mm-hmm. And I'm like, "What?" [laughs] It's like- Multimillion-dollar podcasts- [laughs]...
and then us? That's awesome. Okay. So yeah, thank you.
Thank you, everyone. It's, uh, we will try to keep it entertaining, and, um, we are rambling right now, but we'll keep it on point [laughs] to keep it insightful. So yeah, a couple things I wanted to cover of, uh, cool things that came up at NAB that we didn't talk about before, [laughs] 'cause someone else did say, "I don't even know why I'm at NAB, 'cause I listened to your podcast." [laughs] "And, uh, you kind of already gave everything away, so, like, I already know what's happening here."
[laughs] Those are all compliments, Joey. The work- Your hard work is paying off. The, yes. We're, we're doing the deep research here.
Okay, so a few things that were cool. The one that was, I think, the most standout was from Lightcraft, which, um, Lightcraft, you know, we talked about before. They have the iPhone app- Yeah... that you can m- kind of use as a inexpensive camera tracker.
They've been developing this new platform, platform called Spark, which- Mm-hmm... uh, w- they sort of teased a bit last year, but now they have a prototype. And it solves a lot of the, uh, pre-vis and visualization things that I've wanted to see a solution for. So the idea is it's a web platform.
You can Gaussian splat scan a location, bring in that splat into the, the platform, and then it, you can connect your iPhone or virtual camera and basically have an Unreal Engine Vcam, but all- Mm-hmm... browser based, super easy to use, and you- Nice... can either virtually scout or move your camera around in your browser. It has a tape r- uh, take recorder so you can record camera movements.
You can bring in 3D objects with motion. So they have a demo of, like, a dragon that had an animation built in, and then trigger that and have your camera move and record. You know, do blocking, storyboarding. So you could either use it for pre-vis, or you could use it for, like, a base level gray boxing animation that you would feed to an AI pipeline later on.
Okay, I have some questions. Yeah. One, how responsive is it? In the demo there, and this is also NAB crappy Wi-Fi, it was responsive.
Mm-hmm. Okay. It was like I had the camera, I'm moving it. Like, it felt real time.
So the view of the, uh, r- the real time renderer, the virtual world, do you just slap that iPad onto your physical camera, onto your principal camera? Or do you just use the iPad camera? This is all, like, a pre-vis 3D environment, so it's much more like the Vcam in Unreal Engine. Okay, okay.
So you're using the iPad or your phone through the Lightcraft app as your camera device. You also don't need to do this. You could just, you know, use the camera in the web browser and do all this in the web browser. But it is, like, a very much m- easier to use Unreal Engine system- Yeah, yeah...
with Gaussian splat support... I mean, I've set up VCams with Unreal before. It's gotten easier and easier as, you know, they, Epic- Yeah... has an app and everything.
But then it, all of, a lot of the take set up and, you know, like, uh, file naming convention, and, like, all the professional stuff around it, I think this offers as a add-on on top of that platform. Yeah. I mean, with that you still have to set up Unreal. You need to be the- Exactly...
computer. You need, like, all that stuff. Yeah, it's, it's like- This is all browser based. That's all, there's nothing to install or set up.
It just syncsAnd yeah, you can see in this demo, this is the VCam. It has a tape, take recorder. I can record the camera movement. I can trigger animated elements in the scene.
I could bring in Gaussian splat scans. So just a cool... It was a really cool system. Nice.
They said they're supposed to launch it later this summer. Yeah, the, the virtual camera thing is something I've really wanted of just having, being able to use a phone or some physical device as a virtual camera- Right... in a space, previs or as a, you know, just base layer pipeline to feed into Kling or CDance to style transfer to have, like, AI videos, but with camera control. Yes.
Exactly. So you're, you were thinking, uh, previs with Lightcraft Spark, and then take that reference video and then, quote unquote, "neural render it with AI." Yeah, exactly. Either that or- Okay...
I mean, even for shoots and stuff, I could see... The- There was definitely a bucket of new apps that were web-based Gaussian splat viewers. Like, District Cinema had their own platform too that was like a Frame.io for Gaussian splats.
Mm-hmm. That you- most of them were kinda designed to work with the XGrids PortalCam, where you can scan it. It's very good. Mm-hmm.
Yeah. It's accurate, and you're getting the LIDAR data, so all your measurements are accurate. And some of them had more detailed things for actual previs planning, where it's, like, running cable, and you can figure out like, okay, how much cable do you actually... like, distance do you actually need?
Or if you wanted to bring in physical objects, just getting correct measurements, and if you were actually trying to lay out equipment in your physical space that you're gonna film in, having all of that detailed information with accurate measurements so you could do it virtually on the computer, uh, before you actually go there, and you can make sure everything's already blocked and laid out, and it's, the measurements are precise. Yeah, 100%. Um, my friend Rene, who's at District Cinema, gave me a preview actually a few days ago.
It does solve the problem of laying down and sort of having real-world units and measurements in Gaussian splats. Mm-hmm. So, like, once you have a accurate scan of your set, then, you know, where does the light stand go? Where does the camera go?
Yeah. Where, where do you block, and where do you put the actors, and so on. So, like, it, it does add a layer of usability on top of what is, like, a, a non-editable thing. Yeah.
And, and, and taking advantage of having all of those precise measurements, uh, and putting it to use. Lightcraft Spark was probably one of the cooler things I saw. Let's see. Other thing that was cool, Selects by Cutback.
So when it comes to AI-assisted editing, Eddie AI is probably the one that comes up the most and that we've talked about a bunch. Uh, Selects was another version of AI-assisted editing. Initially s- when he was giving me the demo, it seemed a bit more targeted at podcast editing, YouTube, talking head editing, um, and, like, some vlogging type stuff. Mm-hmm.
But they had a better integration with, uh, Claude and MCP servers. So you could use their system to kinda sort and edit the footage. Um, it had a sort of like an auto B-roll assisted editing feature to, like, lay down B-roll, uh, based on what people are talking about. He gave me an example of, like, a company or production company that c- th- each Selects project has a limit of the amount of footage it could have, but he was like, they just did a bunch of projects, and then the MCP server would just use the reference from each project and automatically...
Kinda using it as, like, a smart footage search, and then building cuts based on with Claude to analyze a bunch of Selects projects and be able to communicate, analyze all the footage, and then put together new edits. That was probably- Okay. Okay, wow... a more convoluted way to explain that.
But it, it was... They had a lot of foundation levels there where, like, on the surface, it was designed for, like, hey, you can edit your podcast quicker, but they are thinking more about this agentic AI workflow. And that was something I was trying to keep an eye out for of, like, what companies are thinking of- For sure... building their tools for ways to communicate with a Claude code or an OpenClau or these other agentic AI things.
And Selects was definitely in that bucket of future thinking of, uh, yeah, it could edit and analyze your footage, but it could also make it accessible to AI agents. Yeah. I think once the highway has been established between an NLE to, uh, agentic, uh, LLM, it, it's only a matter of time before the LLM gets better and better and smarter and smarter and is able to make better and better editing decisions within your NLE. So this, this highway is, is the first step, which seems like they have built, and now it's just a matter of, like, evolving and iterating on it to make it really, really usable.
And it's, a lot of these are starting with some of the sorta easier things to do, like a podcast with multicam, where it's a predictable format. Uh, you know- Right... you're basically just gonna cut out the ums and the ahs and kinda clean it up. It's not the best at, you know, more abstract editing.
That's what Eddie AI does a bit better at that and is a bit more targeted at. Here's a bunch of interviews- Yeah, like creative editing with, like, voiceovers and- Yeah, creating a narrative arc... beautiful B-roll and- With- Yeah... yeah, the hero journey.
Um, Eddie AI is still more tuned to do that stuff, but they're all, there's still a, a long, long way to go. I, I think that's, that's where you still need a human in the loop. For sure. I, I don't think that part of editing, like, editing as an art form is essentially that, and what this is doing is really just cutting back on tedious, repetitive tasks.
Exactly, yeah. And, and just cleaning up those first takes so that you can have a decent first cut to start with and then just start cutting and doing the fine, the, the craft of editing versus syncing angles and cleaning up mistakes and, and stuff like that. All right. What else we got?
Strada we already kinda talked about. I'll, I'll... In the future I'll, um, talk about my experience with it 'cause we used that for the NAB videos. Oh, yeah.
And it m- mostly worked. It's definitely the... It's, the foundation's there. It's very- Dude, you were cranking those videos- [laughs]...
from the show floor, huh? Still have more to- Is Joey on the show floor or he's in a hotel room editing? Still have more. What's going on?
We... It was the magic of the cloud. [laughs] Got remote GPUs. Or is he gambling downstairs?
[laughs] Got remote GPUs. Have all that debt to pay off. [laughs] What else have we got? Oh, this one.
Um, Shade, which I'd seen them before, but they're... It's a cool kind of all-in-one managed system where you could upload your footage. Uh, they have cloud storage, archivable storage, uh, but they basically integrated AI with the entire system. So everything you upload gets analyzed for visual search.
You can mount it onto your computer so it reads like a local drive. So it's sort of like this all-in-one, slightly cheaper than a Sweet Studio or LucidLink. Mm-hmm. It's kind of like a media asset manager plus editing sidekick platform.
Uh, I thought it was very interesting what they're building. It does have the issue where, like, everything you store, you are paying for cloud storage, um- Right... which, you know, that's, that's, that's the issue with, like, that with- Yeah. Like, this, this year and last year really, it seems like the competition has really heated up for cloud-based media workloads- And that you can mount on your computer...
more so than ever before. Yeah. Yeah. Yeah.
Uh, because we have Shade, we've got obviously, obviously LucidLink, Sweet Studios- Yep... and then Frame- Frame Drive. Yeah... Drive.
But- [laughs] Strata... one thing I did find out about Frame, which it was not very clear in the, uh, marketing, was they're gonna have two different storage tiers. When they initially announced this, the impression was, oh, anything you have stored on your Frame.io drive would be able to mount on your computer, and it could read it like a local fire- file.
Right. It's gonna be two different storage tiers. There's gonna be a pricier... They w- they don't know what the price is, but there's gonna be a pricier Frame.
io Drive storage where you pay more per terabyte if you want your stuff to be able to mount onto your computer, and then the whatever standard pricing is now of Frame.io that you already use that is just cloud storage. Okay. You can't mount that on your computer.
Okay. So it will be more money [laughs] to, to use, to use the Drive mountable stuff. They said they would include some storage in every plan. They just don't know what, and they don't know what the per terabyte pricing of the mount storage will be, but, um- Dude, you're, you're, you're not listening to my tin fo- tin foil hat theory, my conspiracy theory.
Last time you totally ignored me. That's okay. What was, what was it? [laughs] I keep telling you that cloud-based services are gonna get more expensive because of AI, because we're just using up more cloud GPUs and storage for AI.
So there's- I mean-... there's only so many GPUs to go around. I believe in the first part of your theory where, like, cloud storage will keep going up. I think it'll just keep going up because they want more money, and you're stuck.
[laughs] No. No, no, no. I, I, I think that they're- 'Cause what are you gonna do? You get, like...
Oh, move off the cloud. What a, like, what a, what a nightmare. You're, like... Stuff's already there, and you're just like, "Ugh, all right.
I guess we'll, I guess we'll pay more money." All right. 'Cause the alternative to that is- This is the second time-... is, like-...
you have ignored my warnings. [laughs] I'm telling you, cloud stuff is gonna get more expensive. I don't disagree with that. I just don't think it's AI related.
I think it's j- It's not because of AI. I think it's because they just, you know, they can't because they can. [laughs] Like- Because they can. Well- [laughs]...
yeah, I think a lot of businesses are doing that. But I think fundamentally, you know, like AWS or Azure sets a price, and then their price is on top of that price, right? So that price is going up. But, I mean, how much...
For cloud storage, I mean, how much GPU stuff's really involved in that? That's not a, that's not a GPU thing. It's not a GPU-intensive task, but it's still using the same network switches and the same sort of, uh, pods and- Uh-huh... containers that, you know, an AI inference, uh, infrastructure would use, right?
So even if it's not, like, low on GPU usage, it's still, um, using the same highway. Side thing, but I started [laughs] listening to a podcast that's a four-hour history of data centers. [laughs] Why did you do that to yourself? [laughs] I had a lot of desert driving over the last few weeks.
[laughs] It's very interesting. It is very interesting. [laughs] Yeah. All right.
I, I don't disagree with the pricing going up. I just think it's because they can, not because of AI. So, uh, quick sidetrack. I was in a LA neighborhood d- um, yesterday, and, um, they were trying to build a data center there I guess 'cause every single front lawn had a no data center sign on it.
Like, they're signing- Ah... some petition. Like, this is, like, a ongoing thing with, you know, can we feed this AI machine enough data centers and GPUs to keep it going? And it's, like, literally up to our front lawn now.
[laughs] With the signs and stuff? Yeah. Like, there's- Yeah... there's actual city council and petitions voting on this, uh- Yeah...
data center that would be, um... It's near a downtown neighborhood over here. Okay. All right.
Yeah. Yeah. They're coming. They're coming.
Until we build them on floating ships or spaceships. And what did I see a picture of? Oh. [laughs] A data center is in a Spirit Airlines airplane.
[laughs] They're like, "Let's, let's turn the Spirit Airlines planes into cer- into, uh, into data centers." I think the wildest one I've seen is, um, data centers in the Middle East, like in, uh, Dubai or some place where it's- Yeah... incredibly hot. Yeah.
And they are trying to cool a data center on top of that. Probably not a good idea. Okay, last one that was pretty cool, Martini, uh, an AI platform that... Look, like all of them, I mean, all do the, kind of the similar things, but their kind of cool unique features are very big and responsive, like, canvas-based editor.
Not node-based editor, canvas editor to, like, lay out all your stuff. Uh, multi-user. People can, multiple people can be in it. But they also have a built-in timeline editor in the same canvas.
So as you're making the shots- Mm... you could do your edit, pull in shots, cut it, and then export and send that to whatever NLE you want to. The... They work with all the models.
You know, they bring everything in. Uh, but the cool thing that they custom-built is this angle editor. So you can basically give it a frame of your film or se- or a thr- AI-generated frame, and then it will- Wow... analyze the people and sort of Gaussian splat it, and then turn it into a 3D space where you can move a 3D camera and reposition- Wow...
and get a new angle, and then have some of the best camera control that I've seen, where it's like- Yeah... you wanna get over the shoulder. You wanna get specific angles. So this was super coolAnd it's just kind of a cool platform.
They, they demoed it for me a co- a, a few weeks ago, and was happy to... I actually didn't even realize they were gonna be there. It was one of those things where I was walking by, and then they were in like in a AI startup spot all the way in the back of West Hall. And then I was like, "Hey, man, like, [laughs] let's, let's talk about this thing, 'cause you're thing, what you're doing is cool."
Yeah. And yeah, this has been one of the most popular NAB videos that we've done, so people are responsive to it. But yeah, I think it's cool- This is nuts... platform.
I, I think it totally makes sense to have, uh, 3D control over 2D generations. And if you go a step further, imagine tying a VCAM, like a Lightspark- Mm-hmm... a li- um, you know, uh- Sorry. Lightcraft- Blanking...
Spark. [laughs] Like a Lightcraft Spark to this system. And so in real time, you have so- like, some level of interactivity. Yeah.
And now you're really getting somewhere with cinematography. Yeah. Yeah, for sure. Yeah.
If [laughs] one of the sides can figure out a way to connect the camera, either Lightcraft has- Yeah... the 3D, uh, turn it into a 3D scene, or Martini figures out how to add your camera to it, that would be a good, a good combo. Yeah. Yeah.
This is a cool app. This is a very cool platform. Very cool. Yeah.
I agree. Yeah. All right. That is highlights of NAB things that I saw that we did not initially talk about.
Yeah. And if you've seen something at NAB that we didn't cover- Yeah... throw us a, throw us a comment, and we'll talk about it on the next- Yeah. The one thing I didn't see that I saw some videos pop up was, uh, someone had built a iPhone bullet time rig that- Mm-hmm...
would do a Gaussian splat, and you could do a bullet time with the iPhone. I didn't see that booth. I don't know where it was. Um, but I saw some videos pop up, so- Mm...
I missed that one. But that, the demos looked cool. You could kick a wall and do a, you know, make it look like you're Neo with the iPhones. Yeah, I would have to see it.
I don't know, man. [laughs] It was a bullet time rig with iPhones. Yeah. So you need like 32 iPhones?
Yeah, you need a lot of iPhones. Yeah. Nobody's gonna do that. They did it on, um, on, uh, 28 Years Later.
Well, like, now you have, uh, star frame and end frame with Kling. You can just do a bullet time- Mm... just have a start frame, end frame, you know? Like, who's gonna buy 32 iPhones?
All right. This'll be an interesting shootout. That'll be an interesting shootout in the future, what gets you better quality. If you have- Well- If you- Obviously, the iPhone's gonna get you better quality, but- For now...
if it's just a one-second cut in your five-minute video- Yeah... are you gonna get 32 iPhones, or are you just gonna use Kling? Depends what you're filming. [laughs] There it is.
I mean, what if, uh, you know, you're trying to do something with liquids or stuff that's, like, kinda detailed and messy? Sure. Sure. If you want, like, highly accurate physical simulation of the world, yeah.
But you know- It makes sense... actually, I was gonna say, oh, sports could use it. But, uh, but, um, I've been seeing more sports demos where they only have a few cameras- Mm-hmm... and they do those 3D camera moves based on a few angles.
They interpolate. Yeah. Yeah, and generate what's in between. So yeah, even sports wouldn't do that.
I don't know. It looks cool, man. It looks cool at the, the show. Yeah.
If I even knew where the booth was. I did not see it. As much as we ran around, I did not, I did not see this booth. Yeah, it would be a cool booth, that for sure, you know.
Yeah, it was a good booth to see it. Just a demo. All right. Let's talk about Claude.
[laughs] Claude, while we were off the air, Claude continued to take over, uh, and sidekick with just seems just about, like, every creative tool at this point. The first one was they kind of announced a whole sprint of, like, partnerships with integration into various creative tools, notably, uh, Adobe Creative crap, Creative Suite. So Claude can integrate with Photoshop, Premiere, uh, just about every tool in Adobe, uh, Creative Suite. That's because, uh, Adobe Creative Suite has a very robust API, so you, you know, for third-party plug-ins and such.
So it was probably an easier task 'cause the, the connections already exist. Yeah, but it's interesting. I mean, do you kn- well, do you know if it can see what you're doing in the project besides... 'Cause I've tried this with, um, Resolve, and you know, it can- Mm-hmm...
it can do things in Resolve, but it doesn't, it doesn't really, it can't really see what it's doing. Like, in order to see, it has to export from Resolve and then analyze it. Oh. Or have the screenshot.
Like, it's not doing live screenshots and constantly- Yeah. So it's like I'm like- Yeah, yeah... "Okay, can you, like, add these videos to this track, or can you-" Mm... "um, add these nodes to this video?"
And it's like, "I can try." And then it's like, "I don't know if I did it 'cause I can't, like, see what I did." Yeah, so it's more of, like, a one-way street where Claude- Yeah... is just instructing the DCC, but it's not really feeding back.
Yeah. And so I am curious. I'll mess around with this with, like, Premiere. I'm curious if this is aware of what it's doing, of what it's able to do.
Yeah. I'm guessing no. Like, the, the live screenshot thing is very, you know, bandwidth intensive and resource intensive. So as a first step, it's just to get Claude talking to the Creative Cloud.
Mm-hmm. And then I think second or third step would be to have, like, a real-time iterative workflow where you're just, like, kinda talking it, talking Claude through your instructions, and then it's doing it on the whim. All right. I'm gonna swing around with Premiere.
But it, I, I, I think this stuff's useful 'cause it's r- it's helpful to just, like, for that base, if you have batch things that you have to do or you're just trying to get- Yeah... projects set up, it, it speeds up tedious stuff. You know, people that freak out, like, "Oh, it's doing all the work for you," like, it definitely is not doing that [laughs] for you. No.
It just speeds up- No... tedious work. You know what is doing the work for you? GBT Image 2.
[laughs] Yeah, if you wanna talk about, uh, graphic design and stuff. But we'll get to it. We'll get to that in a second. The other one, um, there were a bunch of, like, announcements from Claude integrating with a bunch of tools.
Oh, yeah, they integrated with, um, the music, uh, one of the music mixing tools. There were a bunch. The other one was Blender. So it announced a partnership with Blender, and also that Anthropic was going to be a whatever, enterprise, Blender enterprise level sponsor, which means they're given like- Mm...
a couple hundred thousand each year, uh, to support Blender. Yeah. Is, is Blender still, like, open source and- Blender is free and open source. It, like- Right...
exists based on donations and corporate sponsors. So they could use the money. [laughs] Yes. Right.
You would as- yes. So they announced Blender Connector. On top of being a patron of Blender and announcing this enterprise, uh, sponsorship, they have this integrator that, you know, is basically like you can talk to Claude and start having it build out, like, elements and worlds and gray boxing and other things in Blender, which seemsFor us that are in the AI space, very powerful. Uh, not gonna, like, replace a knowledgeable Blender person, but if you are a beginner with Blender or are trying to do some basic stuff and wanna get something built out in Blender, which is a really good 3D tool, seem pretty useful and powerful.
Yeah, right. Um- That makes sense to me. Like, uh, I would welcome a Claude Blender integration just for tedious things like you said, just importing 10 different assets into Blender and then having it write hierarchy, uh, in the outline or something like that. Yeah, I mean, the thing I've always wanted is, like, 'cause, um, you know, I've, I've been m- messing around more with having gray box 3D space as, like, a rough 3D area that we have more control over and can reposition cameras, and then using that as a foundation to give to Clang or CDance, um- Right...
as, like, a, as a motion base layer. You know, and if I could give it instructions of, like, "Hey, I need to build, like, a city street," and just have, like, rough shapes built up to look like a street, that seems like a very good use of Claude to build out the scene. So anyways, the, uh, anti-AI mob grabbed their pitchforks and- Mm... yelled at Blender, and Blender sort of backtracked.
They basically said that they would take the Anthropic donation as a one-off donation and then not continue the annual support. Seems ridiculous. Okay. Their yearly donation of €240,000, which was roughly the cost of four full-time devs.
Maybe I'm not understanding it or maybe you're not either. If any of our viewers know sort of the nuances of why there was so much backlash, please- Why-... throw us a comment. I mean, are you n- no, you're not online as much.
Why? Because people hear AI, and then they instantly, like, get angry. Is it, is it that simple, though? Like, it's 2026.
It's, uh, it, uh, it's literally that. [laughs] "Oh, like, AI's gonna replace all the Blender artists." It's like, no, they're gonna, like, super, uh, you know, empower Blender artists to do- Yeah... bigger and better things.
Yeah, Blender's just gonna get better at doing higher level tasks, like, uh, you know, you know, going from, like, individual modeling of one character to actually building a world with 50 different characters or building a game- Yeah... you know. Also, like, how does this tool stay relevant in this AI future where- Right... so much, you know, it's like, oh, so many things that you would need a Blender for, you don't even need Blender or any 3D tool for that matter.
Yeah. So, you know, how does Blender, which is already open source, stay relevant? You know, and like- I think Blender has a, has a huge place in the future of, uh, generative AI. Yeah, you need that control.
You still need that, you know, at a certain level, you need that fidelity and control- Exactly... and being able to work in a 3D space, and why not bring- And this is a free tool. This is gonna be- Yeah... the most sort of proliferated versus, like, Maya or Unreal, or Unreal's free, too.
But, you know, it's, it's li- it's like the Blender donut, right? Like everybody- Yeah... can do that. Yeah, and then in the process of, like, you know, get everyone their pitchforks, full full-time, four full-time developers are not gonna get hired to, like, continue to work on Blender because the money's, the long-term money was pulled.
Okay. I'm curious how this plays... I mean, they said they'll take the money now, and then they won't do it in the future, like, each year, but also, like, a year is a long time away, especially in AI time. So, like- Yeah...
I would be surprised if maybe they just reverse this in a few months, and then everyone forgot and moved on. Yeah. I, okay, I mean, that, that's, this is a big setback for Blender not taking that money in my opinion. But okay.
No, also they shouldn't have given into this AI- Right. Yeah, yeah... anti-AI mob. Like, you know, it's just...
I, but I, I could also, like, guess that the, uh, the people upset about this have never touched Blender in their life and like [laughs] Yeah, 110%. [laughs] 100%. Like, do you know how hard it is to do anything in Blender? [laughs] Yeah, just hear AI and they're like, "Ah, it's the end."
Like- Okay... yeah... like, please go use Blender and, and, and as a- Yeah... and see if you can, what you can do with it.
Right. A- and, like, just like you said, like the soft- just because Blender is fully built out doesn't mean it need, doesn't need maintenance and debugging and just sort of regular software development things. You need money to kinda keep that going and keep Blender, you know, on the latest Nvidia GPU or the latest Windows or whatever. Like, it's a constant cycle of evolution.
Yeah, and Blender's used at the top level. That Blender, uh, Flo, the Academy Award-winning animated film from two years ago was, like, entirely done on Blender. It is an extremely up-to-date, powerful software that is free to use, and it's only free because of donations and corporate sponsors that help fund it. Exactly.
All right, uh, last couple update, general updates that came out while we were off the air. Uh, biggest one, OpenAI GPT Image 2 officially released. Yeah. What have, have you messed with it?
Yeah, we, we totally missed the entire boat on covering this, uh, so sorry about that, viewers. But I'm sure you've seen other- No, we gave, we gave, we gave the early peek... other places. We said, we, we gave the early, we gave the, uh, the, the early access hint of, uh, oh, hey, look, this new model surfacing, the grip tape.
Yeah, sure. We were, we were, we were ahead of the, we were ahead of the wave. [laughs] We were ahead of the wave, but we didn't let them generate it- Ahead of the wave came, and then we forgot about it... and do side-by-side comparisons.
Maybe we'll do that- We'll do that next time... down the road. We'll do that next time. Yeah.
Okay. Uh, have you messed with it a bit? Yes. What have you found?
So, uh, it's not really built for film and television use. Having said that, it is incredibly powerful for graphic des- design. It's almost like you have your own graphics designer at your disposal, and you just tell this agent, like, "Make me a kid's birthday invite," or, "Make me a restaurant menu," or, "Make me a poster for my work." Make my kids a doctor's note.
[laughs] Come on, dude, just stop talking about the doctor notes. That, their- You brought it up... schools are gonna catch on. [laughs] That's my little secret, and I'm sharing it with the viewers, uh, all 100 of them.
I don't know. [laughs] Okay, so very good at te- I have to, I have to give it a, a tech spin. I, I was trying it with, um- No, no, no, go further than that. Like, okay, imagine- Oh, a graphic design, very good at graphic design...
yeah, like, we're, we're throwing a theater live show for Denoised and make that invite, give it a thumbnail, and see what it does. I'm telling you, it's gonna come out with something pretty cool. All right, we'll try that. We'll try that.
Okay. I will say it wasn't, GPT Image 2 wasn't out yet, but all of our NAB thumbnails, which sort of have this stylized text- Mm-hmm... this textWas generated with, I wanna say it was Not a Banana Pro. And then it was inserted into the screen grabs with another Not a Banana Pro process.
So this whole thing was the first time where, like, it wasn't done with Photoshop, it was done with an AI workflow. You didn't make your head bigger. It, it [laughs] You know, like the thumbnails, like Mr. Beast goes 20% bigger on the head.
We didn't go- Yeah... super stylized with these. Yeah [laughs]. But these were like our style from before.
But, um- Right, right. It's flat. Yeah... I mean, crazy, yeah.
But it, crazy that it is at the point now where it can make edits to images and not mess up the base images, and the text is sharp. Text is fully sharp, fully readable. Text is sharp. Yeah.
The images were, like, edited, but not modeled- I mean-... not changed... were you doing this in the past, like, in Photoshop? Were you just, like- It would be in Photoshop...
putting- Yeah, it would be in Photoshop. Yeah. Like maybe it'll take you about five- Or Illustrator... five minutes a thumbnail, something like that?
It was, like, kinda templated, so yeah. It was, you know- Yeah... um, a, a, a really talented graphic designer. She would make the, kinda pre-build the templates to, like, save time, and then just, like, update the text and insert with the images and ad- adjust the angle so it kinda fits on whatever 3D space makes sense.
Yeah, I'm honestly... She was out of town during NAB, so then it was like, okay, um, uh, Steven, not a graphic designer but good at, uh, AI, like, was able to do this, you know, and not, you know, he doesn't know Photoshop, but he, he knows how to work Not a Banana Pro, and then was able to achieve the same, the same result. Yep. And because the models have gotten good enough to not mess up faces and stuff like they used to.
Exactly. Yeah, I, I don't think GPT-2 would mess up faces either. And it'll actually give you- No... uh, like a level of, God, I hate to say this, but a level of creativity to that text, um, you know.
It just, just, it'll, it'll play with it a little bit more. Yeah, I do. Yeah. Well, GPT-2 came out mid-NAB, which also I was, like, telling people [laughs] at NAB, I'm like, "Oh, hey, did you see GPT-2?
It was finally released." And they're like, "What?" [laughs] So not a lot of people were, at NAB, were ex- as excited about that. Yeah.
Um, but yeah, I'm not gonna have a chance- Your text, your text was like, "Yeah, nobody at NAB cares, Addy. Stop texting me." No one cares. "I'm in Vegas."
[laughs] I'm like, "Uh, Joey, I thought we were friends, man." [laughs] No one, no one cares about the new AI models at, at NAB. [laughs] They, they do. Co- co- co-ed at Martini, he cared.
He knew about it. He cared. Okay, okay. Yeah.
He was the only one that cared. [laughs] There's some AI nerds out there, I know it. Yeah. All right, uh, quick one.
Kling AI now, uh, supports 4K videos. Oh my God. Which is- That's insane... pretty big, 'cause I don't think anyone else really natively supports 4K.
It's all, like, an uprights workflow. Um- Yeah, and from what I can understand, like, this is a huge leap in GPU usage and sort of, um- [laughs]... architecting the entire thing. This is why the storage prices are going up.
[laughs] Yeah, thanks. [laughs] Exactly. [laughs] 'Cause we're going from 2K to 4K now. [laughs] Like we did 20 years ago.
[laughs] Yeah, I, I have not messed with it yet, but, uh, from this demo, you know, just even if you're going down to, to, to HD, having the 4K and then downresing it- Right, downresing is so much useful... uh, you're gonna get that sharpness. I would be curious, 'cause all these demos are of course really good shots, and they're close-ups and detailed. I'd be curious how this 4K works with wide crowd shots.
Wide, right. And wide shots where, like, those are still the trickiest with AI. There's not enough pixels, the details fall apart. I'm gonna have to test it with that, 'cause- Ooh...
the demos, you're obviously, they always pick the best looking shots for the demos. But they're all nice close-ups with sweat on the skin and- I mean, it's getting harder-... moisture... and harder to tell what AI slop is.
And I think the bump to 4K will make it so much more difficult. Then we're gonna be looking at nuances in lighting and, you know, tone mapping and things like that, instead of just, like, sheer- Yeah... uh, like, skin detail missing, things like, obvious things like that. Yeah.
Yeah. E- ex- 100%. And then the last one, uh, just because this started circulating around online. The Academy, from the Academy Awards, has banned AI from ever winning an Oscar.
[laughs] Uh, it basically means- So-... like the AI can't be nominated. [laughs] Like you need a- Well- You need a human to be nominated, which, you know, not really, 'cause we- That, that feels really obvious... like a surprise.
Like- Yeah... it, we're not gonna have, uh, Claude buy a Unitree robot and then come up to the podium to get an Oscar. [laughs] Yeah. Like, that's never gonna happen.
[laughs] It's always gonna be a human that's gonna be using AI to do the thing that they do. And is that okay? Yeah. I'm su- I'm guessing that's fine.
Uh, only roles, uh, demonstrably performed by humans with their consent- Right... can be nominated for Oscars. So this feels more like a flex news drop. Like- Screenplay.
Yeah, it feels like- It does. You know what? We don't, we don't-... a- AI can't come up here...
I don't know, just clarifying something that seemed like a no-brainer. Nah, yeah, but also wanted to mention this 'cause I feel like I heard some people be like, "Oh, I heard, like, you can't use AI at all for, like, the Academy Awards." Like- That's not true... interpreting it that you couldn't use AI, AI at all, like, in your screenplay or, you know, whatever.
A, how would you even prove that? And B, no. It's just like there has to be a human author to be nominated for- Yeah... the thing, which is a no-brainer.
So- I, I think the Academy and, you know, not just the Academy itself, but also, like, the Emmys, the Television Academy, and a, a lot of these government bodies for film and television, I think they will genuinely care about the ethical sort of certified organic, uh, vibe of the actual output, right? Like, what model is it made on? Is it trained ethically? Is it stealing data?
Da, da, da. And then they should, right? That makes sense. But having said that, the fundamental rule is, like, AI is a tool for the creatives- Mm-hmm...
as, you know, and the creative's gonna combine AI with traditional tools, uh, to make the thing that they make. And then it's fine, and that totally tracks with me, I think. Yeah. Yeah.
I mean, it's like, okay, where does that fit in in the evolution of computers being used in visual effects- Exactly... and everything else? It's like another part of that process, so. Yeah, it's like saying computers- Tool evolution...
can't win AI. Well, computers are not gonna w- computers can't win Oscars. Right. I think we brought this up before.
Yeah. But I've been waiting to dig in if it was true, but it was, like, something like the original Tron was disqual- was not, was disqualified from being nominated for VFX because they used computers in the, the VFX, you know, which seems crazy today. Yeah. I alsoDon't know if that's just an internet factor or if that was actually true, but the idea resonates, so we'll go with it's a [laughs] in spirit.
We'll go with it's true in spirit. I don't think anyone was pushing for an AI bot to be qu- to be, uh, able to qualify for- It'll be garbage... an- Nobody will watch it... an Academy Award But if you give AI to, like, a, I don't know, Darren Aronofsky, somebody like that, and they make a masterpiece with it, nobody will really care it's AI.
Like, it's just a damn good movie. I think the more important question is, um, will AI qualify for the Razzies? That's the more important question. That is a very important question.
[laughs] And I hope it does 'cause [laughs] Me too. Or it might just flood and inundate the Razzies, and they'll be like, "No, no, no. No more AI stuff. We want human slop."
[laughs] Right. We want Halle Berry to come here- Yes... and claim- Exactly... her Razzie again.
[laughs] Yeah, and grab it. I think which she did, right? Uh, the other- She did. She was like one of the only people that showed up and got hers.
[laughs] That's awesome. [laughs] That shows character, man. I think it's the same year she won the Academy Award. It was, uh- Right...
what did- Monsters Ball... Catwoman? No. Mon- It was Monsters Ball with the Razzie.
It was for- Monsters Ball and Catwoman, yeah, at the same year. Yeah. Maybe? Yeah.
Yeah. [laughs] That, that's, that's wild. What a swing. [laughs] Yeah.
Um, did you see YouTube has some AI demonetization stuff happening? I, yeah, I have that saved somewhere to put it on this thing. Yeah. Yeah.
Our, it was like, um, like a, like a face detector- Okay... kinda thing. Okay. Yeah.
I think it was, it seemed similar to the music, the, the content license. Yeah, like- I can't think of the, the current name, but-... when it detects licensed music and shuts you down if you're not licensed, yeah... either shuts you down or switches the monetization over to the author.
Right. Um, something like that, but with, uh, actor faces. Mm. I heard there's a way to get- Yeah...
around the music detection thing by just, like, um, expanding the tempo by, like, 2%, so it still sounds the same to us acoustically- [laughs]... but the, the algorithm can't detect it. Or when I see clips pop up on TikTok from movies, but then it has, like, another weird sound, background voice underneath it, and I think it's just a trick or dodge the content- Yeah... moderator from- Totally...
[laughs] finding that it was from a copyright clip. Yeah, I think YouTube realizes- Yeah... that if they don't, like, cap this or just kinda start to police it a little bit, like it'll get really out of hand, and half of all content will be AI slop. [laughs] Well, but that's confusing though because i- you know, on the flip side, they are sort of encouraging that because they have Veo, they have, like, AI generating- Right...
tools built into- Right... more of their tool set. So it's sort of like what do you want? Like what, like- I think they don't want fully automated, just like regurgitation.
Ah. I think V- Right. They wouldn't- But they want some stuff They see Veo as an assistant tool, perhaps not a replacement tool, like most people. I mean, I can understand they would not want some, you know, AI farm just generating random slop and pushing it out to, to like hope that they make some pennies that are more than what the, you know, off of AdSense, that's more than the cost it was to make the stuff.
Yeah. Right. But it provides no value. But they have, yeah, they have these AI tools and these other things built in, so it's not like they don't want any AI on the platform.
They totally, yeah. They, they need it. The fruits, the- The, the bigger company- You know... Alphabet needs AI usage, right?
And what better place for that- Yeah... to proliferate than YouTube? But it can't take over YouTube, so there's like a fine balance. [laughs] Yeah.
I mean, maybe it'll take over Shorts. Uh, you know- Right... I think Shorts is where it's kind of really taken off. Also, YouTube Kids is, is, is insane AI stuff.
Oh, dude, I, I actually- Right out the chute... love this AI YouTube channel called Theoretico, and it's, uh- Okay... I'll, I'll send it to you, and I think some of the viewers know this. It's basically putting a mannequin into different historical places and then what would life be like for you in one week of that time.
So, like, what's life like in the Roman Empire for one week? That sounds cool. And, and then- Yeah... you get time traveled back into your current state, and then you kinda reflect on it.
Well, it was awful, but again, it was very fun, and da, da, da, da. [laughs] And it, it- So it's kinda like a, like a vlog but from like an, a historic period? Yeah, and the reason they're able to do it so well is 'cause how would you shoot that, right? How would you take a mannequin into Roman times and, like, there would be a lot of VFX and animation and so on, but- Oh, yeah.
Yeah. And even the early Y- AI stuff that did well was that POV, like, uh, you wake up on the Titanic or something. Yeah. And, or, and it was just like their POV thing walking through on the ship of the Titanic or through a Roman village.
Yeah. Um, and- I mean, Star Wars Untold- Yeah... is still killing it, Star Wars Stories. Yeah.
Yeah, those, those guy- There's the two Star Wars channels. [laughs] The comment says something like- Well-... [laughs] "Disney execs better be watching this." [laughs] What, what was the name of the channel?
Uh, so, uh, Theoretico is the one that I was talking about. That's it, the skeleton guy. Yeah. There, okay.
Yeah. How, well, they're at 178 subscribers. They're still really small. So, like, it looks like it's mainly- You wake up as a middle class citizen.
Yeah. If you're into history, this- Yeah, it's like a-... this is the jam. Yeah, and it, it's really, like, um, nuanced.
Uh, it, like, it's not all cliche stuff. It's like somebody that did- Mm-hmm... their history research, so you kinda learn from it. Back in our day, all you had was, uh, Wishbone cosplaying as a, as a [laughs] Wishbone...
Roman character. [laughs] Yes. Remember Wishbone- Yeah... the dog?
Yeah, of course. That's a, that's all, that's all we had. Yeah. [laughs] Yeah.
We're living in awesome times. All right. Good place to wrap it up. Good catch-up.
[laughs] Links for everything to talk about at de noisepodcast.com. We've been gone for a minute, but I hope you found this episode useful, and throw us a coup- couple of likes, maybe some high points, eh, eh? High points.
Addy's still on the high point train. [laughs] We'll get you one, Addy. Don't worry. [laughs] Thank you.
All right. Thanks, everyone.