In short
World Labs’ Atlas world model for reconstructing real scenes from image/video inputs, enabling “freeze time” and free camera movement; plus related world-model updates from Runway (GWM World 2, Solaris), and other creator/AI news (local world-model tooling, Gemini video reasoning, Markiplier investing in GoPro and Strata).
Guests
Addy (co-host). No other named guests; the rest are referenced companies/people (e.g., Markiplier, Danny Boyle) rather than interview guests.
Key claims
Atlas blends inputs into a consistent 3D world with plausible logical continuity, letting users reframe and even create synthetic camera POVs for downstream VFX/relighting/stylization/upscaling. It can reconstruct motion accurately (e.g., liquid midair, dominoes mid-fall) from three iPhone videos. Runway’s GWM World 2 improves coherence/quality and supports longer navigation; Solaris generates interactive interfaces in real time.
Notable examples
three iPhones filming a liquid flinging “frozen midair” with camera orbit; dominoes freeze-frame mid-fall; multi-world blending (nature + space station + San Francisco).
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Chapters
Tap a time to open that second in VOThe Power of Reconstruction
0:00 to 0:23
Discussion on the implications of reconstructing the world using technology.
“When it reconstructs the world, it'll build a logical flow that feels like one consistent world that blends.”
The AI Outage
0:34 to 1:19
Conversation about a recent AI outage and its implications.
“Yeah, inference was a little slow this morning.”
Introducing Atlas by World Labs
1:19 to 1:58
Hosts introduce the key news of the week: the Atlas model from World Labs.
“Yeah, I mean, it just goes back to that Cloudflare outages like a few months back.”
Diving into Atlas Features
1:58 to 4:23
In-depth discussion about Atlas's capabilities in video and image processing.
“I would say biggest news this week was Atlas, which is the new world model from World Labs.”
Exploring Creative Possibilities with Atlas
4:23 to 6:15
Discussion on the creative potential and implications of using Atlas.
“you have cameras looking in and then you can just reframe anything.”
Challenges and Future of World Models
6:15 to 8:33
Examination of the challenges and future applications of world models.
“But even if you don't have it and you give it to it, it's not going to look in these tests.”
Runway's New World Model Announcements
8:33 to 14:00
Overview of Runway's new world model and features.
“And then you just place the camera where you need and get the video frames that you want.”
Exploring God Mode in Digital Interfaces
14:00 to 18:03
Discussion on the concept of 'God mode' in digital creation and user interaction.
“You're essentially playing God in this digital space.”
Advancements in 3D Modeling and AI
18:03 to 25:20
A look into the latest developments in AI and 3D modeling technologies and their applications.
“And then like now Runway's like, yeah, we got Gen 5 coming where like the demos looked really good.”
Integration of AI in Google Workspace
25:20 to 28:00
Discussion on how Google is integrating AI into its workspace tools effectively.
“because I'm a fan of AI models that can look at video.”
Show all 17 chapters
Introduction to Markiplier's Impact
28:00 to 28:20
Explore Markiplier's influence in the media industry and his recent ventures.
“and like these formulas and, you know, can figure out these like totals for like a time tracker sheet.”
Markiplier's Film Journey and Success
28:20 to 30:20
Learn about Markiplier's transition from YouTuber to filmmaker and his movie success.
“YouTube and GoPro and the Markiplier stores.”
Markiplier's Camera Review and Insights
30:20 to 31:55
Discuss Markiplier's review of a new GoPro camera and its potential for indie filmmakers.
“And so, and filmmaking fan and has gone through the filmmaking process and now realized, you know, where their pain points and stuff.”
The Evolution of GoPro and Market Competition
31:55 to 34:39
Examine the challenges faced by GoPro and competition from brands like Insta360.
“They really like focused on trying to get, be able to work into professional pipelines.”
Markiplier's Investment in GoPro
34:39 to 36:28
Understand Markiplier's financial interest in GoPro and its implications for the brand.
“So then over the weekend, Bloomberg article came out, but it turns out he invested in GoPro and became one of their largest shareholders.”
Emerging Tech: Strata Investment
36:28 to 38:30
Discover Markiplier's involvement with Strata and its innovative file-sharing solution.
“GoPro also fully supports existing customer products while investing in growth in a broader diversified product roadmap.”
Conclusion and Wrap Up
38:30 to 42:00
Reflect on the discussion and the importance of affordable tools for filmmakers.
“Got a heads up that today he is also investing in Strata, which is the peer-to-peer, plug in your hard drive, turn it into a cloud drive file sharing service slash real-time file workflow service.”
Transcript
Automatic transcript. May contain errors.0:00Joey:When it reconstructs the world, it'll build a logical flow that feels like one consistent world that blends. It just created a logical path that blended it all together.
0:09Addy:There's so much power in somebody's hands to do what they want. I almost feel like we have too much power now across the board to just change anything in anything.
0:22Joey:All right, welcome back to Denoist. Addy, how's it going? Going well, man. How's your morning been? because if you're doing anything with AI, it's been a little bit slow for all of us. Did you encounter this? Oh, really? Yeah, I guess so.
0:36Addy:Yeah, inference was a little slow this morning.
0:39Joey:I was trying to spin up Claude, and then it was like, server's busy. And that's happened before. So then I went to Codex. I think Codex was like, fail. And so we're like, what's happening? Apparently, there's like a big AI outage. I'm also very confused how all the AI. They're all affected by one thing. What is that one thing? I don't know. There was some theories that Atlas, whatever the new open AI crazy model, went online and then decided to just take everyone out. Dude, I mean, I'll believe it, honestly, at this point. Yeah, so they've all been down with their backup again. If you're trying to vibe code this morning, you might have been a little bit slower.
1:19Addy:Yeah, I mean, it just goes back to that Cloudflare outages like a few months back. It's like, wait, we are all pinned down by an Achilles heel that we're not aware of?
1:29Joey:That's not good. Or when there's an AWS outage and then you realize how many sites are on AWS.
1:34Addy:Even like smart appliances at home, like probably your Dyson vacuum probably is connected to AWS.
1:41Joey:Or soon your Dyson toothbrush. Dyson toothbrush. Oh, you do want to talk about it. It has a camera. It's camera related. It's a camera in your mouth. I want one. I don't know. shameful to admit it. I want everything Dyson makes. It's all just crazy expensive. All right. Actual news. I would say biggest news this week was Atlas, which is the new world model from World Labs. This thing's wild. Pretty bonkers.
2:08Addy:Yeah.
2:08Joey:We'll cover some of the demos, but basically similar to Marble-ish, you can give it some images, but you could also give it a couple of video inputs. And then it just creates this from all the demos, very detailed, very accurate 3d model recreation of the real world of videos that you can stop freeze frame create a new camera path around sure it just looks crazy good look at this yeah it looks for the real in some
2:36Addy:of the stuff that has um like actual images as input yeah images as input but the biggest i'd
2:43Joey:say the biggest flex uh especially in our world is this demo everything they've been saying is they shot with three iPhones and it's a dude flinging liquid and it just freezes midair and they can just fly the camera around it. And it's, it's sharp. So the three iPhones are shooting image or video? Video. It's like a 4D Gaussian splat on steroids. I read their article, like you can export a Gaussian splat, but it doesn't, I'm trying to understand like what, what is it capturing? It's not a Gaussian splat. is just generating in real time based on the world model it's generating.
3:22Addy:Yeah, and the detail is incredible too. Like when you showed me this, the first thing my mind went to is like, oh, crap, that's like a phantom slow motion camera. I mean, you can do whatever you want.
3:33Joey:But this is, yeah, so no, to clarify, this is videos. It's three iPhones are shooting video. They gave that as the input. They had some other tests where they gave five iPhone videos. you know obviously the more video the better but from three videos for this for the amount of
3:48Addy:detail crazy i mean it's using a world model under the hood but it's not in the context that you think like when we think world models we think like video games and things like that but like this is reality like it's all underpinned by a world model which leads me into the conversation we were having last pod about uh multiple camera coverage during a shoot and just not really worrying about where the camera is per se.
4:14Joey:Yeah, this is like literally that idea on steroids and now it looks like it's potentially, I mean, these are like 720 output videos you're getting, but you know, that's now, that's today. This is that future where you have a stage, you have cameras looking in and then you can just reframe anything.
4:27Addy:And you can even go a step further, like let's say you create a new synthetic camera POV, you shoot, quote unquote, shoot from this new synthetic camera and then you can take that image and do the rest of your video pipeline with it, like relight it or stylize it or what have you, and then go into an upscaler and then get it cinematic quality. You've just solved for a hardware camera with just a mouse and keyboard. I know.
4:56Joey:And the amount of detail it's able to recreate, and this was an input from a couple images. It also gave examples where you could give it a few images from different worlds, and when it reconstructs the world, it'll build a logical flow that feels like. one consistent world that blends, like I think in this case it was a nature, a space station, and then San Francisco. And that just created a logical path that blended it all together.
5:23Addy:Yeah. It's so much power. It's so much power in somebody's hands to do what they want. I almost feel like we have too much power now across the board to just change anything in anything. Does that make sense?
5:36Joey:Yeah, like there's no constraints. There's no constraints.
5:40Addy:Like, yeah, previously, oh, VFX can't do this. Like you need to scan somebody for that. And now you're like, do I? Like I have a photo of the guy. Like I'll just rebuild him.
5:52Joey:I know. It's like crazy from a couple images. And even before, even with Gaussian splats, it's like if you didn't capture the one side of it, you would have sharpness and that would get fuzzier. And here it's just obviously it's still predicting and it might not be accurate. and it's making it up, but at least whatever it's making it up, it's doing it in a way that feels logical or plausible. I mean, obviously you want accuracy, take more images and give it to it. But even if you don't have it and you give it to it, it's not going to look in these tests. It's not looking weird. It's still there was like a beta sign up.
6:24Joey:So everything we're seeing is from World Labs releasing demos or people at World Labs releasing demos.
6:29Addy:Do you know any idea of like how much time it takes to do things? Like, let's say you give it the three iPhone videos.
6:36Joey:Like, what's the turnaround time? Good question. I don't remember. It didn't say, but I think, I don't think it's too long. Check out this one. Dominoes falling.
6:44Addy:And it just freeze frames.
6:46Joey:Look at that. Oh, wow. Like mid-fall, it freeze frames. And this is from, what, three iPhones, I think, again? That's crazy.
6:54Addy:Well, just like freezing a frame of that fall in mid-fall, all the dominoes in relation to each other, and it's all accurate. it. My God, it's incredible. I don't even know what it's going to be used for because it's almost way ahead of its time.
7:10Joey:I mean, a couple of use cases too that it's giving. Like in one aspect, you have this like video playback and you're able to like freeze. I think it could process up to 60 seconds. Maybe it was less. Play the video, move the camera around in the video playback. But then the other application is you have images, you just give it or you can, it has its text to world prompt building and then you just build a world like this looks i've messed around with marble like i was always like it was okay kind of like if you didn't need to move the camera too much it was okay but this looks like it just solves so many of the problems with yeah where it's just like too fuzzy too sharp right uh or too too fuzzy and not sharp enough to to do much with this does that the other thing that my mind goes to is like we you and i
7:55Addy:I have been seeing kind of an uptick, a rise of a lot of like what I would call traditional 3D controls for video generation. Right. It's like, you know, you bring in an environment or Gaussian splat, then it's a 3D turntable. You rotate it, scale it, whatever, put your camera here and now you can keyframe your camera like it's doing a dolly in or what have you. and then you feed that into a video model and it'll generate and it'll adhere to your camera language that you dictated earlier. This bypasses that. It's almost like you don't even need 3D environment anymore because you're in a world model.
8:37Addy:You're in the real world, if you will. And then you just place the camera where you need and get the video frames that you want.
8:45Joey:Yeah, if we're just talking about purely building a 3D world, Yeah. If you make a couple images that is the world you want and you give it to this and it builds that 3D world, depending on what you're trying to film and do, like, yeah, it could be, it's persistent. Could be good enough. Totally.
9:00Addy:Yeah, exactly. If it's 100 % the thing that you're trying to capture is the world model object and what have you, then you're good to go. The 3D turntable control, what I would call, quote, unquote, like the Unreal Engine type control, That may be just a stopgap for now until we get to a fully realized world model video generation control.
9:26Joey:Slightly tangent but related. I don't know if you saw the new Danny Boyle trailer for Ink. No, I have not. His new film, but it had this one shot. Yeah, come on. That's cling. Come on, Danny. No, I think that was like a couple cameras around the table. They did a bullet time, yeah. Poor man bullet time. But now I'm thinking like, oh, man, you just give him. Wait until he discovers Atlas.
9:55Addy:I wouldn't be surprised if it was like sparse keyframes into a video model just for that shot. Just like, yeah, get one in there. I think you're overthinking this. I think it's literally like four cameras. Do you think so? But it's so fluid and smooth.
10:09Joey:It wasn't that fluid. It was like pretty jittery.
10:12Addy:Oh, okay. Yeah.
10:14Joey:All right. Fair enough. Okay. Yeah. Yeah. I think we're too much in AI land. Poor man's bullet time. I did this years ago. We were trying to do bullet time, but we didn't have enough money. But we had a bunch of flip cameras. And so we arranged flip cameras on an array and then filmed stuff blowing up and did poor man bullet time.
10:35Addy:Yeah, I guess as long as you take from each camera and then advance time so it's like...
Read the full transcript
10:41Joey:Yeah, just think all your videos up and then you kind of mess around with it however you want. Yeah. All right, so Atlas, while this was all over the internet. Hey, shout out to World Labs. You guys are killing it. Yeah, man, this is crazy. Excited to mess around with this once the beta opens up.
10:56Addy:Yeah.
10:56Joey:Okay, other news. Big week in just world models. Runway's got their own world model announcement.
11:03Addy:An update to their previous world model.
11:06Joey:Yes. This one, GWM World 2, general world model. What did you see with this, Addy?
11:13Addy:Yeah, so to me it feels like it has just longer coherence, so you can indefinitely navigate in this world. I'm not sure what the upper time limit is, but I think the last one was like tens of seconds or something like that, and this one feels much longer.
11:30Joey:It seemed like when I was watching the video, it seemed like once you made the world, it was indefinite. Like you could just stay in it. I think it was like every time you do a new action, it's like regenerating. And then so it sounded like it was persistent. This was just from the explainer video. So it wasn't 100 % clear. But it sounded like it would keep, every action you did was recorded and then it would kind of generate, like keep going with a new branch of the reality. Sure.
12:01Addy:The other big thing that I noticed was just the overall bump in quality, right? Like if you go back to GWM1 demo videos, after a while you start to lose texture and things just start to have like a Gaussian filter on them. This one is able to retain like high frequency details much, much better.
12:22Joey:It's also making sound. It also has like dedicated camera control. Yeah. I mean, pretty cool.
12:27Addy:My question is like, they're obviously going to build it to monetize it. What is it going to be monetized for? Is it just games, do you think? I don't think that's enough.
12:35Joey:I mean, I think all these world models are like, you know, I think we've talked about this before. The real advantage is robotics, industrial. That's where the money is, is training military killer robots. So the Terminator can understand the world. Yeah, urban warfare. Even the original Terminator didn't have vision this good in the original movie. It was all like pixel art. Come on. Let's not touch on the classics here. I think still ultimately it's robotics.
13:06Addy:Yeah, you think that's where the revenue is? I really think there is a huge potential for mobile gaming that is on device.
13:18Joey:A lot of this inference becomes small enough to then just run on your iPhone or whatnot.
13:23Addy:And I guess the challenge is how do you build it in a deterministic way so the level layout is always the same and you always have the level goals sort of laid out per your game action and so on. And yeah, I guess this is just the first step in just being able to do it. And then the next thing would be to add that control to it where you have deterministic outcome.
13:47Joey:Yeah. I mean, I think this also still ties into that vision from that Cristobal video. You know, it was like maybe it was a year ago at this point where he was talking and generating the world and navigating it in real time. And I still think that is an accurate vision of where things will go in this sort of world creation space where you're just directing and stuff is generating and changing and applying in real time. Yeah.
14:11Addy:You're essentially playing God in this digital space. Yeah. God mode. God mode. The other runway announcement, did you want to cover that?
14:20Joey:Yeah, yeah. So this was the other one that I saw earlier. So this one's called Solaris. Yeah. And it's what they're calling their first interface world model. and this is an operating system that generates interactive interfaces in real time. Yeah, this is really streaming the interface.
14:36Addy:Really thinking out of the box here. Kudos to Runway for just churning their brain in the direction that a lot of people are not. This goes back to being in God mode, right? So if you are in a world model and you want to do things in the world model, then what is the interface for it? How does the user actually, what are the mechanics of the user experience like? And I think this is that. If you look at some of the demos, it's like, oh yeah, that's not like clicking buttons. You're like moving things around in the world, but it's intuitive enough for you to do it.
15:11Joey:Yeah, and imagine if this is on the web or like they showcase some home decor examples and stuff and it's like on the web, you're moving around products. Incredible.
15:19Addy:Picture you're living for a moment.
15:20Joey:It's like actually updating in real time. Yeah, that's crazy.
15:24Addy:But wait a minute, Addy.
15:26Joey:So you're down with this generative interface. But when we talked about that Gemini preview of like a fluid web websites that were generated on the fly, you're like, I don't get that, man. I don't get that.
15:42Addy:But this, you're like, generative interfaces. I'm all on board with this. Yes, yes, yes, yes. It's just more elegant. It's just more user-friendly. That one gives me like synthetic AI overlord vibes. Like, do I need a synthetic website here? And so on. This one is like, I'm already committed. I'm already in the world model. Shoot, might as well just play God in here. What's the best interface for me to do that?
16:08Joey:Well, I think what they're saying is like, this is something that would, you know, you're not in the runway zone. This is like a, I don't know, like a standard, like a web standard.
16:18Addy:Yeah, that's yet to be seen, right? Like how do they actually productize it and make it universal across other types of media or applications?
16:29Joey:Yeah, it seems like something that could be deployed as like a back end to other companies, other interfaces and stuff.
16:36Addy:These are so open-ended, man. Everything we're covering here, right? Like the World Labs thing, this, the Runways model, like we don't even know, or if our viewers do, please comment.
16:47Joey:I think in a lot of cases they don't even know. It's like we have this tech, we're doing this thing. They don't fully know how it's going to be applied or where people are going to find uses for it. It's just like you can do this. We think it could be useful for this, but there could be someone who finds it.
17:03Addy:Also, just thinking about this from a runway leadership perspective, building things that have kind of like a tunnel that you've just opened up with the light at the end will attract more investment and make more money hopefully in the future for them as they continue to build these things. So they don't necessarily have to dictate the clear path to market adoption or monetization and all the nitty-gritty business things. They just have to be like, hey, this thing could change the world. All user interface, as you know it, could change. All right, investors are like, okay, I'm putting in another billion dollars.
17:44Addy:Here you go. Run away. So I think there's a lot of that as well. It's like you don't really have to build for a thing today. You just keep building.
17:53Joey:Yeah, and if you build the standard that other stuff's built on. Yeah, sky's the limit. Good place to be. Runway kind of, you know, hadn't heard much from them. They'd be here about Sea Dance and other stuff. And then like now Runway's like, yeah, we got Gen 5 coming where like the demos looked really good. We got this new world model. We got this new interface model.
18:11Addy:Yeah, I like Runway. I mean, my touchpoint with Brunway the most is Aleph 2. Like, that's a solid video-to-video model. It has multiple keyframes. Like, you can get a lot of stuff out of that. So it's nice to see other things out of them as well.
18:25Joey:Other news. McMuppets had a version, and I've seen, I don't know if they just turned this into a version, but, like, basically taking Minimax H3 and building out some workflows that turned it into a, it is not going to be as high fidelity as what we've just seen, but a world model builder that could run locally on your computer. That's cool. That's very cool. Using MiniMacs and stuff. Yeah, so I've seen a couple from this one as a new model on Hugging Face. I think it seems like it's built off of what the pipeline McMuppets had developed.
18:56Addy:Yeah, McMuppets has been posting a lot of that oral model on Comfy stuff. Shout out to him. That's very cool.
19:03Joey:Yeah, and this is stuff that I had saved as interesting and tied into the 3D space, but I have not fully looked into it yet. But Hyper 3D, World Gen, this one looks like image to a 3D mesh model. Sure. Which that I'm always interested in, especially like if it gives us something that we could work with in Blender and 3D space. Yeah, like if it's modeled correctly.
19:24Addy:Yeah.
19:24Joey:But I mean, also so many of the updates from the newer models like Fable 5.1. I've seen some crazy demos where it's just building out stuff in Blender. GLM, the Chinese company, their model, their new Flash model built this like Blender model that was like way more detailed than our Culver Theater model and that all just random Blender. So there's just been some crazy updates with all these new models that are very good at 3D and CAD build outs.
19:54Addy:I don't know how it does it, but ever since the Amazon Culver Theater thing that you did with Codex, I'm just like, man, we're not going to need to model or rig or shade
20:06Joey:or anything in the future. You just kind of ask and you shall receive. Yeah, it's kind of crazy. There is one theory on why they might have gotten better at this. I saw this blurb that there was some news that came out that basically the reason there had been such a shortage on Mac minis was not because everyone decided to make their own open claw, but because OpenAI, the usual suspects were buying them all up.
20:34Addy:Yeah, and shout out to some of our viewers who commented and educated us on MLX. The MLX platform on a Mac is kind of like CUDA where it can parallel process AI tasks like a GPU.
20:51Joey:Interesting. Okay, yeah, there were some good comments on that episode of what people have run on Mac midis.
20:56Addy:Yeah, and I really appreciate that. Our audience is super smart. It's smarter than us. I love that. The other thing they told me that was new to me and you, I think, is Hugging Face makes money not just from hosting the weights, but also hosting the inference of those weights on their platform. So imagine AWS for your open weights. You can host it on Hugging Face.
21:21Joey:Yeah, that was interesting too. And that makes sense. You have your custom model. You want to compute. it still doesn't seem like why it would be worth 13 billion because I think Nvidia has their own hosting that offers that as well. Oh really? I think so. I mean yeah it makes sense. I don't know if it's 13 billion dollars since but I think it's a factor of all the things we talked about.
21:43Addy:Yeah for sure.
21:44Joey:It's appealing to Nvidia to have it and also to protect it. I know they just posted a statement about why they're buying it not why they're buying it but they're buying it and that they remain committed to open source models and they're better. You just don't know which way the future is going to go. It's probably both and they know it. I think for them, it's obviously for them, we talked about this, it's better. It's like the more open models and stuff that people need to run on hardware, their hardware, the better it is for them.
22:10Addy:It's like Linux, right? Like Linux is like 90 % of the world's data centers and like appliances run on Linux. But then you and I never use Linux. We use Windows or Mac because consumers don't use Linux much. But then there's a huge need for Linux. Hardcore.
22:28Joey:Yeah. Unless you're hardcore.
22:30Addy:So it's the same way. I think in the future, closed models will be for guys like you and me who are just trying to get something done. But then the open-weight models is what everything else runs on because it's just easily modifiable and adjustable and so on.
22:47Joey:Yeah, I think when you start going to scale and you're like, we know what you got to do and we got to do a bunch of it then figuring out like building out your own server rack or something and running stuff locally starts making sense exactly kind of dabbling around and stuff it's tougher especially with like how quickly models and everything keep improving and changing yep yeah so anyways uh open i bought a bunch of mac minis
23:11Addy:all of them by a bunch how many like millions i don't know probably whatever i don't know if
23:17Joey:actual number. But yeah, basically it was to have macOS running all the time and run tasks on it to train the models so you get better at computer use, specifically macOS computer use.
23:30Addy:So macOS minis at OpenAI are doing the management for the inference or the training?
23:37Joey:No, I think it's literally AI is running the mac minis to just have real life just computer use cases. Oh, so they're simulating
23:47Addy:users with Mac Minis?
23:49Joey:And I guess it's cheaper to just buy the Mac Mini than it is to buy like cloud compute, cloud... Oh, like pods, virtual machines? Run pods or something? Like virtual machines? Yeah. Or like a virtual Mac Mini. VMs. Or virtual Mac machines. Yeah. I guess it's just it was cheaper to just buy the computer and run it 24-7 then. Which probably that would make sense to buy a$600 computer and run it 24-7 then to pay for the virtual machines.
24:14Addy:That's interesting. So the Mac is the consumer, if you will. Instead of hiring people on Fiverr or whatnot, they're just buying Macs.
24:26Joey:Right, buying Macs and just running, I don't even know, stuff all the time to create training data on how to use a Mac.
24:31Addy:Yeah, okay, that makes sense.
24:32Joey:You know what I'm bummed about
24:34Addy:is Apple is raising prices across the board now. Sounds like the new CEO's first day was yesterday or something. John Ternus?
24:43Joey:Yeah.
24:44Addy:And yeah, I guess everything is going to go up in price because of the shortage of silicon across the board.
24:51Joey:I thought they announced that a while ago. I mean, because of just the rising silicon and the rising cost of solid-state drives and just everything that's going up.
24:59Addy:Yeah, everything, yeah.
25:00Joey:Yeah. I don't think I was saying that was like a month or two ago. Okay. Even when that new Mac Mini, when the new Mac Minis were announced, like the base price was like$599, but now the new base price is$799. Something like that, yeah. 600 to 800 books. So yeah, there's no more$600 Mac minis. Dang. For now. One quick thing I wanted to mention because I'm a fan of AI models that can look at video. So Google's been kind of pushing these like smaller updates with Gemini, like their Flash models. There was like a 3.7 Flash and now 3.8 Flash. But 3.7 Flash can now do a Gentic video. So you can give video as an input and it can do better video reasoning.
25:44Joey:We kind of talked about Kimmy K3 that could do this a month or two ago. But so I'm testing out because on our clipping workflows to see if this performs better or works better. I'm always a fan of AI models that can take in video and analyze the video. So I'm going to see how this does. Oh, interesting.
26:01Addy:Yeah.
26:01Joey:Yeah, that's a specific call-out feature that they have in this new model.
26:04Addy:It's interesting. Google is building a lot of these flash models instead of building a brand new foundational model because by now we would have expected Nano Banana 3 or something to come out, right? Like just periodically.
26:16Joey:Yeah, like I still think it's the pro model is like Gemini 3.5 and they've been, I don't know if it's just because the flashes are quicker and they just want to stay, keep stuff coming until they finish four
26:29Addy:or whatever the new big model is going to be. They're also like value engineering their inference down. So like you don't need as many credits
26:37Joey:to run the flash models, I think, right?
26:39Addy:Yeah, I mean, the Flash ones are cheaper and faster. So maybe they see that the market is now credit conscious, token conscious, and they're like, oh, we should offer stuff at the lower end too just to kind of keep them using and keep them coming back.
26:54Joey:Yeah, I mean, this is the model you're using, whether you realize it or not, if you're doing anything in Gemini or I'm assuming Google Search is running some form of one of these Flash models when you get the AI search results back. And yeah, I mean, Google is like something that's at such a massive scale that they need, you know. They need efficiency and everything. Right. Like it's so many queries that are getting AI results, but like you can't, these are not paying people. So you can't spend so much on those results.
27:26Addy:For sure. But I mean, side note, like the cleanest integration of AI into like day-to-day work that I've seen is what Google has done with their workspace. like putting Nana Banana into Slides and Gemini into Spreadsheet. And like, it's just done so elegantly.
27:43Joey:Yeah, and it's finally, I know there was like a lot of beta stuff and it was like, there were a lot of walls before where you had Gemini in the Google Docs sidebar, but it wouldn't actually do anything to your doc when they first came out. But now it's like finally, like it's been really helpful in Google Sheets where it's like, hey, I need to like make a sheet that like has these columns and like these formulas and, you know, can figure out these like totals for like a time tracker sheet. Yeah. And then it's like, okay, here's the plan. Do you want it? I'm like, yep. And then it like builds out this nice sheet.
28:09Joey:And it's like, cool. That's exactly what like this should have done a while ago, but it does it now and it does a really good job. Yeah, it has more autonomy
28:16Addy:and control over the stuff. Yeah.
28:19Joey:All right, last thing. Let's talk about Markiplier. YouTube and GoPro and the Markiplier stores. This is wild, man.
28:26Addy:I mean, it just goes to show where the media industry is now. A YouTuber is really making some big moves here. So you want to cover what has happened so far?
28:38Joey:All right. So Markiplier, big YouTuber. I don't follow him, but I believe he covers kind of the gaming space, but is also like a filmmaker creator. He got kind of more mainstream notice last year because he produced and self-funded the movie Iron Lung. I think it was a couple million dollar budget. He funded it himself. and then he distributed it mostly himself to theaters. He got all of his fans to call up the theaters, help book the theaters. He worked with the distributor. I love it. But then, and this was sort of pre-obsession and back rooms and that sort of took all of the box office headlines.
29:14Joey:But, or maybe it came out this year. I don't know. My whole timeframe is blurred. Oh, it came out beginning of this year. Yeah. But it did like 20, 30 million in the box office and it was like a big success for like, hey, YouTube guy makes a movie, gets a theatrical run kind of like still controls a lot of the pipeline and you know does really well so it was a good movie then yeah it was based on a game that I think he's covered and so he knew there was an audience and interest for the game because there was a lot of response on his channel so it was already sort of had some validation because he's already built out this whole audience online and so yeah a lot of it was like you know I think we're going to see a lot more of this but like you know YouTube creator creator kind of builds audience, validates IP, and then develops the IP themselves.
29:59Joey:And produces it himself. Through the process. Produces himself, you know, gets a distribution partner that's like 50-50. It's a better, different deal than if you traditionally got a distributor. And controls a lot of the steps along the way that traditionally independent filmmakers give up this kind of control. That's beautiful, man.
30:18Addy:I mean, just carving his own path along the entire route. Yeah. That's awesome. Yeah, it's an awesome success story.
30:24Joey:Uh, so now he's got some cash. And so, uh, and he's a big camera fan. And so, and filmmaking fan and has gone through the filmmaking process and now realized, you know, where their pain points and stuff. I didn't watch the video, but he posted a video like last week that was like a review hype video of the new GoPro camera. That's like a micro four third sensor. Uh, I can take interchangeable lenses. This is like their new push and what they're doing. And he compared it to the red and was like, Hey, this is a really good camera, especially if you're a film indie filmmaker it's pretty affordable you can make really good stuff with it it has really good quality you know comparable to like a seven thousand dollar red and you liked
31:01Addy:it um it probably shoots raw and everything so that that makes it even more wonderful yeah it
31:06Joey:was like 8k raw open gate had a lot of you know i mean it's basically a sensor in a box but for i think it's 700 like it seems pretty good i personally think it's a weird take for gopro to but um um like if you're looking for a good camera sensor it seemed viable you think his his thought
31:29Addy:process was coming from like producing iron long and just seeing how that was shot is like wait most of that could have been done with a camera way cheaper than what we were using possibly or
31:42Joey:just like his audience was like i want to make stuff too but like i don't have five grand seven grand to drop on a red camera and it's like what can i do and it's like well gopro has a has a new camera that's pretty affordable and like is punching above its weight class and what it
31:55Addy:gopro has always punched above its weight class so i mean it just it just sucks that they're no longer in the center of the camera universe anymore but i mean you remember like 10 15 years ago it was it was all a rage yeah yeah i mean they were advanced because they were small cameras but
32:11Joey:they had really good color science. They really like focused on trying to get, be able to work into professional pipelines.
32:18Addy:They were great cameras,
32:19Joey:but then they, I think probably, I mean, I don't know. You know what? They, but they went public.
32:25Addy:They acquired Cineform, I think. Viewers, correct me if I'm wrong. Oh, that's right. Cineform was this like codec that was as good as ProRes, or if not better. Like it was an incredibly high quality codec to put on a tiny little camera. and that gave GoPro such a big leap up in quality, even though the center was small and everything was small.
32:47Joey:Business-wise, they went public. That probably brought in more pressure. They did Karma, the drone thing. That was a flop. I think that took a lot of resources. Their stock evaporated overnight.
33:01Addy:I have some friends who used to work at GoPro during the golden years, and they just kind of told me, like yep just saw the whole stock thing slip away right underneath us they're also like they had
33:11Joey:such a good adventure brand yeah like they had some of the best action they're kind of like red bull in that way like they were red bull they were like the red bull cameras yeah they like really had that and then it kind of faded but also i mean insta 360 is just killing it they're killing it they're killing it and from a just like pure like product development side everything they've made has like solved so many of the pain points that I've had with GoPro. Their 360 cameras are great.
33:40Addy:I also love DJI's like the DJI Osmo is incredible. The Ronin 4D, which is that crazy hybrid Rialto type camera. Like DJI is also in this space as well. Yeah, everything DJI is making is really good.
33:53Joey:Yeah, instead of 360, like the GoPro made a 360 camera. I've used it. The software is like atrocious. It is, they have like, and it doesn't seem like they've really developed, like continued to work on it. the Insta360s yeah the Insta360 cameras are great they've like replaced any use case I had for GoPro because you can put it and then you're just capturing everything and then you just reframe what you want later and then their GoPro competitor is like it looks like a GoPro but the little camera pod can pop out and it's magnetic you can stick it anywhere and then the screen still works you can flip the screen up all of the things with GoPro you're like what am I looking at like they just designed and solved so many of the pain points.
34:34Joey:Anyways, tangent. But GoPro started it all. GoPro started it all. MarketPlyer is a fan. Invested. So then over the weekend, Bloomberg article came out, but it turns out he invested in GoPro and became one of their largest shareholders. Now, he did invest in, I forgot the two distinctions, but basically whatever the public class of investment was. So it was just like, here's money, I bought stock, but no influencer control in the company, which is like a different class of investment. So it's just like...
35:05Addy:Okay, so it's like just him going out with a pocket full of cash on like Vanguard or Robinhood. Yeah, it's like him going on, yeah, exactly, like E-Trade or whatever and being like,
35:14Joey:hey, just buy me like a million shares of GoPro or whatever. I don't know what it was. So it wasn't anything that gave him like special access or whatever or privileged information.
35:24Addy:Yeah, it goes to show like he really believes in the company. Yeah, he was just like, I really like the company.
35:30Joey:He's like, I think the name brand, he's like, I think it's undervalued. I think the name brand alone is -
35:34Addy:Yeah, he said it's the Kleenex of cameras. I remember that video. Yeah, exactly.
35:38Joey:Yeah, on Philip DeFranco. There was an interview with him on the Philip DeFranco show. The name brand is the Kleenex of cameras. And then like the day after, then it's announced that GoPro is going to be acquired.
35:51Addy:285 million. Wow, that's quite a bit of change for GoPro.
35:54Joey:Well, look, I mean, so they're being acquired. What is the company?
35:57Addy:It's an optical, so they make glass?
36:00Joey:Starman optical. Basically, from the stuff I've seen, GoPro still makes images. I think they still have like a US, I think they have US capabilities to make sensors or lenses or other stuff. And there's a, you know, defense play here, like defense contractor.
36:15Addy:I knew it. I knew it. I was just going to ask you, is this for drones? I think there's like a tie in with that.
36:23Joey:Now, apparently, GoPro will still remain a separate company and still public, like, you know, facing brand. GoPro also fully supports existing customer products while investing in growth in a broader diversified product roadmap.
36:35Addy:Mm-hmm. That means military stuff, basically.
36:39Joey:And especially if they have U.S. or other production capabilities. Yeah.
36:44Addy:Yeah.
36:45Joey:They are building a manufacturing facility in Jersey.
36:48Addy:Yeah. For, like, defense-type deployment, like, everything has to be built on U.S. soil, right? Like, you can't have parts from China coming in or anything like that.
36:57Joey:Trying to probably not. I mean, I'm sure other countries, yeah.
37:00Addy:Yeah, I don't know, maybe. I mean, I don't think it's possible to make everything.
37:05Joey:Yeah, I'm sure it's very controlled, but I don't think it's possible to make every single thing in the US that's needed. I guess it depends on what type of thing for what application, right? So yeah, so that happened. And then there was stuff like, oh, did you know anything? But I don't think you do anything. He says you don't know anything. I believe him. You believe him?
37:23Addy:Yeah.
37:24Joey:I don't know.
37:24Addy:He seems like a nice guy.
37:25Joey:believe him i don't know i believe him i don't i think this is like i think he so like uh i don't
37:32Addy:obviously don't watch his channel much um from what you were telling me today seems like an avid film aficionado like a like a tarantino type if you will right like he loves film he loves cinema he went out of his way to produce an independent movie which i mean i've been there it's painful like even to make a bad movie takes all your energy let alone a good movie right and uh through that process he got into cameras and hardware and then realized like hey um i maybe i have a duty or an obligation to unlock a lot of new filmmakers i'm gonna go preach and then he preaches about gopro and then he invests in it and then this happens after is that right yeah like any the stocks he
38:17Joey:bought were just regular stocks. There wasn't any controlled information or controlled type of investment. Yeah, I think it's just weird timing. So he wasn't just investing in GoPro. Got a heads up that today he is also investing in Strata, which is the peer-to-peer, plug in your hard drive, turn it into a cloud drive file sharing service slash real-time file workflow service.
38:46Addy:We've covered Strata here in the past. It's like a headless, cloudless collaboration tool for editors, right?
38:53Joey:Yeah, or just anyone that needs to deal with files of any size. It's like a lucid link, but without the cloud. So you plug in a hard drive, you could turn it into a cloud drive, and then someone else could access the files and work with them depending on the file size and stuff in real time or kind of near real time. You can pin and cache local files. The stuff they cracked is crazy. Because it's also like, you just pay the monthly fee and there's no size, there's no storage, there's no cloud storage limit. So they really solved a big pain point and he's interested in solving pain points now for creators.
39:29Joey:Actually, funny side story, at NAB because I do a bunch of videos, Justin, the camera guy, we were filming at Strada and then Justin, my camera guy, kind of branched off and I saw him take a selfie with some guy, and then he came back later, and I was like, who was that? He's like, that was Markiplier.
39:50Addy:Oh. He was at AI on the lot? NAB. He was at the Strada booth at NAB. Oh, okay, got it. Okay, so this conversation has been happening for a while.
40:03Joey:Or maybe that was when he discovered it,
40:05Addy:and he's like, what is this magical thing? I mean, I'm going to put money that Markiplier... Markiplier's not done investing. There's a pattern here that's emerging. You have the camera sensor. You have the infrastructure, which is Strata. What's next? Maybe there's an AI tool or something else. Who knows?
40:27Joey:Yeah, or I think you went through this whole process and he's like, oh, there's a lot of pain points in here for being like, if you want to be a filmmaker, low-budget filmmaker. Right, a YouTuber. A budget filmmaker. Who's making products that solve these problems and like where can I invest or put resources behind it?
40:43Addy:I mean, I love it because we tend to be so highbrow, right, in the industry like, oh, IMAX 70 millimeter, right? And just everything that we talk about, think about is always like six figures, seven figures, eight figures. Like we also need stuff at the bottom where people can actually enter and do things, you know, with very, very little money. So it's really good. It's a good move.
41:10Joey:Yeah, I think it's a smart move. I'm more bullish on the Strada investment than the GoPro investment. Yeah, although like - Well, except in the span of a week, that GoPro investment might be paid off and done, so that's good. Exactly, yeah.
41:22Addy:That GoPro investment will then pay for other investments for him. Mm-hmm. You know, yeah. Rather than like us seeing a new camera, I don't think that'll happen.
41:31Joey:Good place to wrap it up. Good place to wrap it up.
41:33Addy:Hey, viewers, thank you again for your comments on the last video. we learned about hugging face and Mac GPUs all from you so thank you and please enlighten us again and to the person who asked if we're always like this yes we are always like this I love that comment it was like are these guys high or this is normal behavior
41:53Joey:this is normal behavior
41:55Addy:why what's up I'm on the opposite of high
41:58Joey:I'm like just drinking coffee during the show so yeah links for everything we talked about at denoisepodcast.com Thanks for watching. We'll catch you in the next episode.
From the publisher
In this episode of Denoised, Joey and Addy break down the latest breakthroughs in AI world models, from World Labs’ photorealistic Atlas to Runway’s GWM World 2 and Solaris interface. They also dive into tech news, including OpenAI’s Mac Mini hoarding ...
