Is Seedream the Next Nano Banana? Plus Comfy Cloud and more film tech updates

0:00Welcome back to the noise. We're going to do our roundup of top AI stories filmmakers.

0:18Let's going to i'm in la we should meet up we should meet up and do a podcast you know distance here is still so long so far away but uh yeah also i'm heading to atlanta tomorrow and then uh then we'll be back together next week yeah that's cool yeah uh one week i'm there one week you're there and then it's been a bouncy summer i really miss our in-person sessions i know it's been a minute looking at you on a teleprompter screen is not the same effect i know our jokes don't hit quite Right. Yeah, that too. With the lag. All right. So we're doing our early, it's earlier in the week. We usually do the roundup on Friday, but we're going to give a shot, try it out earlier in the week.

0:58And also we didn't do it last week. So there we are. Don't kill us, please. Yeah. First up, already people calling it a Nano Banana Killer. Nano Banana has already been usurped. That's crazy. One of my favorite models. I've talked about it a bunch from ByteDance. They came out with SeaDream 4.0. And among other things, it has similar adherence to text prompt modifications where you can tell it what you want to change. And it changes it. Change the angle, change an object, change a thing. You can give it up to six input reference images. It can generate 2K images or up to 4K images, which is also very nice because these look really sharp and really good.

1:37Nano Banana is an image generator, but C Dream 4.0 is a video generator. And they're being compared to each other. No, you're thinking of C Dance. C Dance. cdream is their image generator cdance is the video generator both from byte dance educated well i mean up to 4k resolution i will give a shout out because i usually just use their stuff on file or like through the eight um through apis or through other tools they also have a whole other like their own ai website which we've never talked about because it was sort of not on my radar but it's called dreaming ai that's a good name i like that i've heard it compared to some other things but yeah i learned more about it uh because um jennifer by chance was at tiff at ai on the lot.

2:15And so she educated me too, because I was like, oh, I just used it through APIs. She's like, oh, you just use it through the website as well. So if you are also wondering how to use these models, Dreamina is like the main spot, or I use it on FreePick because it's there. Dreamina is an AI platform. It's not a model, right? Dreamina is their quote FreePick or their like Google Flow. It's like their web portal. It's ByteDance's web portal to use their models among their models being in a bunch of other things but that's their like website portal if you just want to be like i just want to use by dance models you can go to trimuna and use them there gotcha gotcha so i wonder if your tiktok account works with you know getting free stuff good question i mean i think you could also use these in you know cap cut probably i don't know what the integration's like that's a little outside of my world but i yeah i mean by dance is making these models i'm sure they're going to be in every ByteDance product at some point.

3:12Well, hold on. Let's take a step back. I mean, we talked about this before on a few podcast episodes ago. You know, ByteDance owns TikTok, and they're also making these incredible models, and they're not charging anybody for it. They're free. I mean, what is the end game here, Joey? What do you think? You're thinking of Alibaba. Our seed dream. Wait. They're not free. They're not free. No.

3:37We are AI experts. Edit this out. It's been a minute. I haven't podcasted in a while. I'm the one recovering from COVID. There was tube strikes, man. I had to walk all over London. Maybe we should just keep it on just for laughs. I think it's pretty. Blooper's real. I think it's also funny to show. We live in this stuff, and it's so complicated to keep track of. Seriously. It's my fault. I generalize. I overgeneralize too much, so I just thought Chinese models are free. I actually, I do understand. I just had this quick thought once before, too, and then I was like, oh, wait, no, no, no. It's Juan from Alibaba.

4:19That's the free download. Do whatever you want with it. Yeah. The ByteDance models are not free. They are pretty inexpensive, but they're not free and they're not open source. You can't download them and run them on your own system. You have to use it through an API or on Freemita. It's going full circle. Okay. So, I mean, you use it on FreePick because FreePick API connects to Dreamina, I'm sure. Yeah, wherever ByteDance is selling the API access, I use it on FreePick because I got, you know, as we talked about, I, you know, they got me with the ProPlus plan. So it's unlimited. So, you know, we could just generate and I don't worry about the credits and stuff.

4:54It's a good selling point because I'm using it a bunch. Anyway, full circle. Talk about that. Come back to C Dream 4.0. Yeah. Have you tested that against NanoBanam? Yeah. So I have. One, coming back to the updates of this, the other unique thing that it does is you could also tell it to generate multiple images in your output at the same time. So if you're like, you give it one source image of like, you know, us or in the studio, and then you say generate four images of different closeup angles in this scene, it will spit out four images, independent images in that batch generation of, you know, whatever the hand, I don't pick the closeups or whatever you specify in your prompt, which is interesting because in theory that would get you a more consistent output because it's in that same generation.

5:43And that's kind of the same photo quote hack that they did with the video model. See, see dance, see dance, right? With see dance was the video model where they realized if you prompt shot changes in your prompt, you can get a more consistent output of multiple shots in the same generation because it's in that same space. It's sort of this, the photo hack equivalent of that you can do in C dream now. So I think that could also be interesting to get like batches of really, really consistent outputs of, uh, you know, character consistency or location and consistency. If you're trying to get a different angles or poses or details of something.

6:20Is it just four or can you dial it up all the way to like eight or 16? I think it might be eight. I'm just thinking from like a previous perspective, a lot of times you're doing so much trial and error, right? Like you don't even know what you really want. You're just kind of going with the flow a little bit. Like you have some creative, but you're really waiting for the right visual to come along. And in those instances, I would think like putting like 16 little thumbnails in front of you and you pick one or two. I mean, that's a pretty cool feature. Yeah, that is a cool feature. Oh, look at you running Comfy.

6:52Yeah, so I'll also shout out that Comfy did add it as a native API node into their system. Oh, so that dollar sign thing up above, that's the API key? What is that? That indicates that the node costs money and that it'll take it from your account. Usually, if it's a consistent rate, it'll tell you how much it costs per generation. But because it depends how many outputs you pick, the price can vary. But I mean, this was pretty cheap. It was maybe like, I forgot, two cents or four cents a generation. I mean, it's so cool to see the resolution at which you're generating. Those are big resolutions for, you know, AI.

7:26Yeah, exactly. We're typically at like 1024. Exactly. Yeah. Yeah, these have been good. So here, I cranked up max images. Let me go up to 15. Oh, nice. So I guess 15. Yeah. Are you running Comfy Cloud? What are you? Yeah, yeah. This is API. So it just shoots into the cloud and then it brings it back here. Okay. now going back the original question now to banana sea dance comparisons that's the million dollar question that's the question i was messing around with both and it was literally like shot to shot or depending what i needed sometimes now to banana did it sometimes sea dance did it would you say it's 50 50 yeah uh i would say not a banana still for what i was using it for needed it still came out ahead like now to banana i found was way better at 3d space like if i gave it an image of a you know, character in a space.

8:13And I'm like, Oh, you know, I need an over the shoulder shot of this character in this space. The nano banana output felt much more consistent, like as if it was over the shoulder and whatever we saw in the distance here, it felt consistent, the geography of the layout felt consistent. Sea dance didn't quite do it correctly, it kind of shifted a lot of stuff around and changed a lot of things. And then also even I was trying to do a, I wasn't even sure if this would work, I had like a screenshot of a computer screen, like a website, and I added it as an overlay image, you know, in FreePic because I have a little editor to mark up what you want.

8:46So I put it on a AI generated computer screen of like a real computer screenshot. And then I said, you know, blend the computer screen into, blend the screenshot into the computer screen. And I was like wondering, like how is it going to do with the details and stuff? Seadream really warped it so that you really couldn't read it. But Nada Banana looked sharp. Like it kept, you could tell what it was. It kept a lot of the details. I was surprised. It sounds like you, I mean, you won't say it. You're too nice. But it sounds like Nano Banana is still king. I think it's still king. Yeah. I mean, for the stuff I was using it for, that's why I'm like, I didn't give it a big, wide-ranging shot.

9:24But the thing that did come in handy was the multiple outputs. And it was for a thing where I'm just like, I'm trying to ideate and storyboard and think of stuff. And having the multiple outputs and a consistent thing, space, that was cool. So that's cool to see. wouldn't you be able to do that with nano banana like just in a comfy ui just have a bunch of parallel api notes and just generate but you're generating it with different seeds and different seeds right generations like this is in the same latent space the same you're absolutely right yeah so you're in that same imaginary world that is pulling these eight images or up to 15 images from you're right that that could be powerful for certain things when you just need the same consistency across right like a character image and it's like or put this character in like four different locations and it would theoretically be the same character because it's the same or also i'm imagining for training like an ai model you need a bunch of synthetic data and the synthetic data has to line up correctly maybe in those instances oh like if you're trying to generate training data like if you're going to train another ai model on the likeness of a synthetic human, you want to generate that synthetic human in one shot.

10:34So their features are consistent for 15 images. Yeah, that could work too. All right. Next update, veed.io. They've got their own talking avatar model now, veed fabric 1.0, AI talking video model. But basically give it an image, you give it the audio, and it can generate up to a minute long videos faster and cheaper. yeah i mean looking at the preview that you're showing here i mean that look that's as good as like hey gen or maybe even slightly better to be honest i'm not sure like how modifiable these are are these presets or can you just upload anybody's photo and then have i think you just give it any image you know with a character i i don't know because i've had these issues before like how much it of a face it needs to understand like what to map out because like in these demos it It seems like they're all human faces.

11:28Sometimes they do animals. If it's like a dog or like a popular animal that has a very clear face, like sometimes if you get more abstract animals, they don't know what to do with it. But I think the use case for this is more, you know, kind of like in the UGC or explainer video world, like the HeyGen world. Yeah, I was an OG Veed user for a long time. I've been using it for like almost three years. I used to do a lot of social media posts and I would just upload stuff onto my MacBook and then just edit it on Veed, which is a browser-based editing suite. And then the best part was it had really good AI audio cleanup and it also had captions.

12:08A few years ago, it was hard to just generate captions. Now you can find it everywhere. Everything does it. Yeah, and I could see them integrating Avatars. So if you're a content creator who makes a bunch of short-form content and then you just need a B-roll or somebody holding some product or just saying something profound for like five seconds could be a random person. It doesn't matter. I think this could be really handy. They're in the social media tool set that you're talking about. They're kind of in the same space as Descript. Transcripts, text-based editing, can create social clips. Descript, this reminded me though, because Descript, they've been adding a lot of AI editing assisting features, but they also have an AI, I think they can, they had a feature called overdub where you could like modify or tweak or change the audio.

12:58You know, if someone, if you need to change how someone said something with, you get their consent, there's like a thing you have to permission thing you have to read. But I think they also added a video modification feature. So like if we were adding this podcast and like, it's like, oh, you messed up a thing or I messed up a word and I needed to change my mouth moving. I think they added a feature that does that, but it's not their model. I'm assuming it sends it off to Hedra or HN's API or one of those other companies. That's pretty powerful still. I mean, man, it's just like, what can you not do?

13:30Yeah, especially when it comes to editing and you're just like, oh, jump cuts are gone. Like if you and I did this podcast and we had this third character who we just kind of, hey, Joey says this, Addy says this, what do you think? And we cut to this third person, and it could be a totally synthetic AI avatar. But you can totally build that into the story, you know? Yeah, I mean, I think, I'm sure there are podcasts that exist where it's like, yeah, the co-host is like an AI chatbot or something. Yeah, again, I don't know how I would connect with a synthetic podcaster. like so much of i mean you listen to so many podcasts as do i and i then the number one reason we go back for it is we build this familiarity and this um comfort level with the podcaster right yeah yeah like and it's a very emotional human thing and i don't know if that could be replicated with an avatar yeah and on the flip side there's like also some podcasters where uh you know i was like i just really like their style and i'm like i i can't listen like maybe i kind of like the topic but i'm just like ah i just don't like listening to to this yeah you know who comes to mind robin hood the stock trading company called snacks okay and there's two guys that are just way too excited about stocks and pnl and those things and it's like today rivian lost four percent yeah i can see it relax i'm just waking up exactly i think i flagged it maybe we'll cover it next week but there's that story that came out about how there's like one company that's churning out like thousands of podcast ai podcasts and i saw that at a time and um basically just like flooding the market to just try to get any like hit and stuff yeah yeah i got thoughts about that i don't think that's great that's my thought okay we agree on something okay uh cup i thought i had something else to say about veed but i don't remember yeah veed is great go use it oh i mean my general my general note about veed and what it does it's still a bit frustrating that basically every ai avatar model where it's like you want to make an image talk it is give it an image give it an audio track and then it goes and maybe you can give it some sort of performance direction but runway is still the only game in town that can take human performance as the driving force and take the human performance, modify the audio if you want, and then drive the performance of the image.

16:13And so I would love to see if there's more development of that space of being able to actually have a human performance drive the AI creation instead of just... Yeah, I mean, you can even call it animation. I mean, I'm sure animators will... we'll be really pissed that you called it that but yeah, it's puppeteering is what it is there's gotta be a huge market for that we tried to do this a few years ago a friend of mine and I built this animated character in Unreal as a really cute cat, very furry and we had a real-time Unreal very modifiable session like there is triggers and all types of cool stuff and it's trained to twitch live and so he was able to either game and then you know react to the game live or he would watch the news and react to it i mean if if you didn't do it live and you did it uh with ai and then you recorded it and then you just streamed it on twitch after like once it was all done i mean that could still be a viable content that could still be a viable content strategy.

17:29I mean, there's also, there's like Adobe character animator that it's not very good. Oh no. Okay. Yeah. Yeah. I mean, it's, it's flash based or whatever version and they're calling flash. Now it's, it's 2d, right? Right. I know it's 2d. I didn't realize flash still existed in some form. Yeah. They absorbed it into this technology. They don't call it that anymore, but it's essentially that. And 2D animation has its task and its purpose, but with all of the AI puppeteering stuff now and stylization and how easy it is to achieve 2D animated look. Yeah, but I was going to also say, look at just Snapchat and the face filters.

18:11Those are real-time. They look really good. That's possible. You could do it on those. Yeah, I mean, imagine those on steroids. I'm not even asking for real-time. I just want to be able to give it the video. yeah like i can wait five minutes i can wait 10 minutes i don't need you know just give me back something really uh like vfx quality all right next update comfy drops cloud we talked about this a couple episodes back when we were coming remember the comfy episode we did how image generation works and then we did a second one how ai pipelines work and our version of an ai pipeline at the time at the end of that episode i asked you hey joey how does comfy make their money we've wondered this multiple times on the show i think they're figuring it out i think this thing will make them a little bit of money yeah so they're launching comfy cloud basically you could run comfy ui in the cloud this solves each problem because the problem with comfy ui right now is for where to run it effectively you need to have a very powerful computer uh with a very powerful nvidia card and a decent amount of storage space because all these models are take are 20 to 30 gigabytes a model plus everything else you need more i mean the video models can be even bigger than that the problem with running comfy at home is even if you have a 5090 nvidia rtx gpu which is like state-of-the-art you can't get elastic performance which means like if you want to go up to like two gps three gpus and then come back down to one when you're idling you can't really do that at home, right, obviously.

19:47And these backends run H100 GPUs, which are, like, far more powerful. Probably the best GPU we have to date for AI development. And you get access to that. So it's anything, I mean, it's way better than what you would be running at home. The company of the cloud is not new. There are other third parties that do this as a service. You know, I'm curious how this is going to shake up because it's a wait list right now and there's really no word on pricing or how this is going to work or what options you're going to have and how the storage aspect plays. But yeah, I mean, this is the next logical step and we'll open up a big market for Comfy UI users.

20:27Yeah, I mean, there are sort of AI infrastructure, browser-based AI infrastructure services like Replicate comes to mind. That's another one. But the thing that Replicate is missing is like a node-based developer UI, which Comfy obviously has. but there's also Weavey, there's also Invoke, there's also Flora, there's also LTX. Yeah, we're talking about those, but you can't do as crazy of a workflow as you can with Comfy. You can't build your own node. You probably don't have access to a bunch of other people's nodes that Comfy has. Comfy's amassed a big library of nodes that have been pre-built.

21:09Yeah, nodes to do things. I mean, even look at the McMuppets videos we talked about where it's like, oh, look, you can also pose track, you can camera track, you can create masks in Comfy. You can do all the nodes that other people built. Other people have built, exactly, because it's all open. And Comfy is even good at doing non-AI functions. So the other day I saw somebody use a color LUT in Comfy. Oh, interesting. Yeah, so you can do color transforms and stuff in Comfy as well. Yeah, I mean, I've used it just, like you said, that's kind of just a good integrator or orchestrator for building out some quick nodes or just simple workflows.

21:47I mean, the nice thing I do like about running a company locally is everything you generate, you can save directly to your computer. So if you're sort of doing a bunch of batch work or processing, it is nice where it's just like, oh, the file's already there and I'll have to try to redownload them or something. But yeah, I mean, I'm sure maybe they'll have cloud connections or something that could just sync it to a folder on your drive. Yeah, it's saying local and cloud feature sync. I mean, I'd imagine it would be like DaVinci Resolve running on Blackmagic Cloud as well as things locally. Yeah, something that syncs.

22:16Or maybe like a LucidLink kind of sync or something. I'm sure it'll take a couple of versions to get everything right. But this is definitely a step in the right direction. Yeah, for sure. I mean, I would expect their pricing to be fair slash on the low end. And with all the API stuff, the nodes that they built in, they don't do any markup. They just buy credit and they just charge you whatever the APIs are charging them directly. So I'm kind of curious how the price is going to come out. But I would expect it to be pretty accessible. Yeah, the only caveat there is AWS is not cheap, no matter what.

22:53So even if they pass the cost to you, you're still looking at anywhere from, I'm just guesstimating here. $0.04 to$0.08 an image and maybe$0.10 to$0.20 a video. If you're running something that's like a model that you would be running locally, like stability or the local flux models. Like a 1.2. Yeah. Yeah, exactly. I'm excited to see it come out. Speaking of stability. There we go. Best to just talk about this one. Yeah, quick caveat. So as we always say on the podcast, this is not sponsored by my employer or Joey's employers. We do an independent podcast. With that said, we're covering the stuff that I do at work.

23:39Yeah, this is a big deal for stability because stable audio 2.5, I think it really becomes useful for professional enterprise users in a big way because it's all trained on licensed music. And the stuff that it generates, I think you own, you'd have to double check with stability on that. But you have full rights to all your generations. And then the cool thing is, because it's trained on licensed music data, it is kind of dialed in to make audio for your YouTube video or your documentary. Any kind of background music, any kind of... I need to just fill this out. I need to just lay down the tracks.

24:17This is the tool to do it. The generation's super fast. I think a 30-second track takes like three seconds to generate. Yeah, that's fast. It's actually so fast. it can almost run on your computer and then the best part is um you can actually go in and you edit you can edit the middle intro or outro it's called audio and painting kind of like image and painting i'm not sure if they launched the audio to audio yet but so let's say like like the adobe example i remember when adobe had the foley thing where you can go there you're and it fills it yeah So this has the same audio feature as well. That's cool.

24:54But this one, unlike sort of the other stability models, you can't run this locally? Like you have to run it through an API or a cloud? I'm going to say you can load it at stableaudio.com or through the stability API? For now, it's through the API. I'm hoping it just is downloadable and open source like every other stable stability products. Could it just be like a smaller version of the model or something that could run locally? Again, I don't work on the audio stuff much, but I'm going to say this is light enough to run on an iPhone locally. Yeah, it's like pretty fast, dude. I'm going to have to give this one a shot.

25:33Because yeah, just that need for like just background beat, something to fill in the background as always. Yeah, and again, like this is my sort of personal opinion on it, not the companies or anything. I think it is pretty one-to-one with Eleven Labs' solution. like as far as quality goes and okay how usable it is for a professional like it's not it's one thing for me to make like some instagram reel and add music to that that's it's not i don't think it's a big deal right like but if you joey are making a documentary and you need to have really like finite control over the generation and the beats per minute and the instruments used and the cadence and the tempo like this is built by musicians so it speaks musician prompts like it understands what a crescendo is and arpeggio and all the words that i don't quite understand cool but that's also always a challenge for me when i'm trying to prompt the music thing because i'm like i want that thing but i don't know what it's actually called in music terms right right yeah i'm like i'm just like you but then this is built for like proper musicians that's cool well i'm playing the video that's probably not going to do anything because i'm not listening to the audio and i'm just seeing a giant cat and jump in the screen yeah but yeah i'll check it out and mess around with it and we can uh maybe maybe we'll try some like background beats for the podcast yeah let's play the 50 cent song all right camera updates because i love this and cameras new red camera yeah yeah i'm a big fan of red you know this yes um i also love my sony venices and aria lexas like i just like high-end cameras there's so much uh art and science behind how these cameras are made and how they're used and what a cinematographer can do with it.

27:11I love the world of this stuff. So, V-Raptor is now available as a new version called XE. It is smaller. It's only four pounds. It's compact, I would say. It's the same size, maybe a little bit bigger than a Komodo. Okay. And best of all, it's cheaper. So, now it's coming in just under$15 ,000 with the same performance. The Raptor before was, I think, north of$22 ,000, somewhere in the$25 ,000 range. And then it went up to V-Raptor XL, which is even more expensive, bigger body, and so on. Yeah, so I think V-Raptors now come in three flavors. You have the big, I call it the Zack Snyder one, the one that he always, with the curved handle in the back.

27:57That's the big boy. And then you have V-Raptor Regular, which is about$20 ,000 or so,$25 ,000. for the body and now they have the light via raptor light which is 15 000 so this is um you know if you're looking at the sony lineup you have the fx9 uh fx6 and fx3 so i think uh red is just really trying to accommodate all the different price points and all the different size formats and different markets really like i said the big bad boy is for feature film where you know there's a lot of money to build expensive rigs and get techno cranes and all that stuff right but most of the time you don't have that kind of budget so you're running and gunning with a much smaller camera one that is you know you can rig it up with small rig or tilt up parts and go from there and that's the v-raptor regular and then what if you want to put the camera on a drone or what if you want to put the camera on the side of a car with suction cups then you need to go much much lighter And the problem with the Komodo is I think the sensor is not as performant as the V-Raptor sensor.

29:03So the V-Raptor has better low light performance, obviously has 8K where the Komodo is 6K. And this guy can do really nice slow motion stuff. I think the Komodo, Komodo, I don't think goes beyond 60. I might do 120, but with the V-Raptor, you're going 240, you know, so. Yeah, 240 at 2K. Yeah, that's awesome. Yeah, yeah. So you might be wondering why RED blended the V-Raptor with the Komodo as far as size goes and everything. I think they're two totally different cameras, and V-Raptor is up here as far as professional quality goes. So this is a good move on their part. Yeah, it's like you get the V-Raptor as the A-cam, and the Komodo as your B-cam or backup cam for the cut-together.

29:46Yeah, I mean, the color science would be similar, but again, the V-Raptor is really, if you shoot stuff with the ViewRaptor and you shoot stuff with the Komodo, you look at it side by side, the Komodo does struggle I know my friends I read are going to hate me for saying this, but the Komodo does struggle with lack of light. It's really grainy and it just shows up even in the most modest circumstances like where an iPhone wouldn't show any noise a Komodo does. Is that where the Raptor excels because they highlighted it as a future? Yeah, that sensor is just something else. It's a different engineering underneath and the best part is these now have z mounts on them which is nikon's mount so you know nikon one can argue is more famous for its glass its lenses than for its cameras so now you have all the legendary it's non-existent film cameras or cameras moving making moves in the film space absolutely yes that's why they acquired red very slow to the game on that one yeah yeah so this is a yeah nikon glass is legendary you know nikon's been making glass for i don't know 50 years or so so you all those glass is now easily adoptable for the red but speaking of nikon and cameras they also have their first camera since acquiring red the nikon zr yeah so this is right and then uh this is actually a twin announcement because canon just announced the uh i believe the r50 c50 c50 the c50 which is a smaller version of the r5c which is like already really solid camera so you're playing in the i'm gonna say like in the five to ten thousand dollar range for body you can factor me on that i think i think you're right on this one yeah so uh in this category you have the sony fx3 you have the nikon zr and you have the canon c50 and these three are in the same shape and size of a dslr so you can see that's like you can grip that with like one hand but because they're quote-unquote professional cameras you can gen lock them you can you know even even less 2200 so those are i mean we should shoot the podcast with like three of these look at this nikon red yeah yeah yeah and it's starting to get like the red aesthetic so you see you see if you go to the image again in the back it has like the camphored uh edges on the top like uh and then it has the white text on the black plastic yeah that's all red aesthetics it's pretty cool yeah and it looks pretty pretty like like you can drop that thing and nothing will happen to it looks a little bit more robust i mean it's literally just a sensor flip screen yeah so one of the biggest challenges with the canon c5 i've used one extensively was the fan because these things have to operate you know in any temperature any humidity they have big fans on them and when that fan goes off they like the audio gets picked up like that noise so when it records the fan shuts off and then the sensor overheats so this is this is one of those things that is not a big deal for the bigger body cameras because the fan is so far away and it it's bigger so it doesn't have to spin as fast and so it's got red not red log right red color science yeah red color science would include a 12-bit raw okay yeah so it's red raw so yeah the red raw format the r3d format has the red native color space embedded in it and of course you can take it to any color space and post yeah 12 bit raw log g310 i mean that's a 12 bit raw 6k that's awesome and crazy for uh internal recording for a 2200 tiny camera large amount of 32 bit float audio which is always great to have and then oh does it have built in i don't know yeah if you get it if you get this nx mobile air attachment you could do frame i o camera to cloud built in the camera oh wow that's awesome that's cool yeah the reds uh so red's built out three different v-raptors and they have two different komodos they have the regular komodo and the komodo the bigger one the big boy and now they have i would i would call this a red camera yeah they have the nikon zr so i guess this is a new label z cinema so this kind of i'm guessing more kind of consumer friendly version of the red technology tied into the nikon brand it's you know what it is it's probably the red color science and the red compression on the raw format So those are the two IPs that Nikon was probably after when they acquired them.

34:09Yeah, it's still crazy that they had such a lockdown on raw patents. Yeah, I mean, they were one of the first, right? Like the Red One. Was it the Red One, the very first Red camera? I mean, that was a game changer during this time. All right, cool. It was fun to cover camera stuff. Hey, Nikon, hey, Red, just send us a couple cameras. We'll test them out. We'll test them out and never return them. Long-term test. Long-term test. All right, and more just kind of interesting business updates. Vimeo has been acquired by Bending Spoons for$1.38 billion. Oh, wow. Bending Spoons, the garbage bin of tech companies.

34:47I called them a vacuum cleaner. Vacuum cleaner, that's much better. A hoover of distressed tech companies. Who else do they own? Bending Spoons owns Evernote, which, I mean, basically they own all the companies that I was very fond of in the early Web 2.0 days. they own Evernote they own WeTransfer which I also did not realize until I read the press the release. They're buying Vimeo which Vimeo, as I think we have talked about, has been like, what are they doing? What are they all about? Because yeah, I think going way, way back to early days of web video, when YouTube came out, it was YouTube and it was Vimeo and they're pretty equal and when you were like someone in full school, Vimeo was like the place where you put your film because their web player was way better quality It still is like that most artists that I know have their reel on Vimeo.