Tech Lust, Smart Upgrades, LLMs, And The Modern Color Suite

Analog Life And Human Craft

SPEAKER_02 0:00

You would think that teenagers would wholly embrace this stuff. My kids, to a large part, themselves as well as their friend groups are all about that analog life. I mean, like, literally to the point of being like, hey, do you think I could get like an old like walkman on eBay to listen to like tapes? And I'm just like, uh, what? Right? We feel it as that generation that's kind of bridged that gap of growing up analog, moving into computers and digital. I think the kids, they can't describe what's going on about it. And so the human part of them is naturally grafting on to just like when we were teenagers or whatever, like people were like, hey, collecting vinyls, cool, right? I think that same thing is gonna be throughout this age of AI, where people like, you're gonna go to a facility who's like, oh wow, you have real human colorists? Like, we want to work with you guys because there's humans pushing the buttons.

SPEAKER_01 0:57

Welcome to Color and Coffee, a podcast that's focused on the craft of color and the artists behind it. I'm your host, Jason Bodak, and each episode we'll sit down with some of the most talented artists in the industry and have a casual chat from one artist to another. We'll share their stories, their insights, their tips, and maybe even a little gear talk. Whether you're a seasoned pro or just getting started, join us for some great color discussion. Sit back, relax, you're listening to Color and Coffee. Hello and welcome to the Color and Coffee podcast. I'm thrilled to have you for this episode, and I have really, really special guests on this episode. I have Joey Deanna and Robbie Carmen, the hosts of the Offset Podcast and partners from DC Color. Welcome to the show, guys.

SPEAKER_02 1:52

Hey, bud, how's it going?

SPEAKER_01 1:55

Thank you so much for joining us. I feel like we're combining the the two color podcasts finally.

SPEAKER_02 2:00

Yeah, I feel I'm honored as well because I think if I'm not mistaken, this makes a return to your podcast for both Joey and I. We uh return our participants, so excited about that as well.

SPEAKER_01 2:12

Yes, both of you are returning, and I'm glad we got to do like a round table kind of thing going on here. So I'm really excited to have both you guys on that we can just chat about some fun stuff. And one of the first questions that I want to ask both of you guys is first, if you're not aware, Joey and Robbie run a really fantastic podcast called the Offset Podcast, and they recently hit their 50th episode. Be sure to check it out if you aren't already. And I want to ask you guys over 50 full episodes. What is for each of you one of your favorite topics that you guys have discussed so far?

Favorite Topics From 50 Episodes

SPEAKER_02 2:45

Joey, you want to hit that one first? I was gonna say, why don't you start? Okay, just like recording, Rob, you you started off, right? Okay. Um, yeah, well, first of all, thanks, Jason, for having us on. It's always good to touch base with you and talk to you. And we love your podcast as well. So thanks for the kind words. Really appreciate that. We do these pretty frequently. So it's once every twice a month, every two weeks, basically, hit you know, plus or minus a day or two. So at a certain point, they kind of blend together a little bit. But if I had to say, like some of my general favorite topics are some of like the soft ones, like the more emotive kind of stuff, right? That kind of thing. Um uh recently, uh maybe end of last year in the fall, we did one about managing relationships in post-production, which I thought was which was really good. Sort of talking a little bit about both clients, but also like, you know, dealing with your bosses, your colleagues, you know, that kind of stuff. Another one sort of like that, that I thought was kind of hit with the same thread I like is sort of things they don't tell you about being a colorist. That was about a year ago, sort of like the schlag sometimes of having to deal with constant changes and waiting for changes. And it's so I don't know, those are fun. And then from a technical point of view, one of my more recent favorite ones was we did a show on virtualization, and that was exciting because literally since like October of last year, it's pretty much all I've been thinking about is virtualization and building out new VM servers and stuff like that. I mean, just this week, still doing the same thing. So we tend to do things that are kind of like pertinent and sort of live to our, you know, our current work life situation. So if you ever see one that's like, you know, Rob thinks about quitting the industry, it's probably just that's what's happening this week, you know, that kind of thing.

SPEAKER_01 4:26

Love it, man.

SPEAKER_00 4:27

Yeah. For me, I really like the workflow-related topics because that's where I kind of one have the most opinions, and two can really go into some detail, and I think really help people. I'd say probably my magnum opus of our podcasts has been the two-part series that we did on how to deal with archival sources, both from film and from old video sources, because it's a topic that I think has gone really neglected by the greater community. And I feel like watching even the most highest budget documentaries that are out there now, some of the older footage and archival footage has been, for lack of a better term, just destroyed. People don't take as much care as they used to. And I don't think they're trying to take less care. I think as the industry has marched on and we've gotten further and further away from these older formats, the knowledge has gone away. People don't know where they're going wrong. And I think hopefully we can help a couple, we can help kind of raise the bar a little bit for documentaries because both Robbie and I have been in the documentary world for most of our entire careers. So it's kind of personal to me when I watch a documentary on a topic topic that I'm like really into and I'm really liking the documentary, and then I'm like, why are there scan lines all over this and why is your emotions stuttering? And did did anybody watch this before you put it on TV? You know, so I think with that one, one, we went into a lot of depth, and I think we did a really nice job, so I'm really proud of it. But also, I think we can really help some people raise the bar on their product, and that's kind of a good feeling. I hope it helps people.

SPEAKER_02 6:03

And I would just add one more thing. It's that one of the challenges that I'm sure you face were on your show too, right? Is that I have a heavy education background. I've written a lot of books, done a lot of uh in-person training, uh online tutorial stuff. Lynda.com has started a color-focused training website a number of years ago. Anyway, I think one of the hard things about doing a for us, because we are both very verbose technical, talkative guys, is that balance act between doing like tutorial and like chatting about a subject, right? Sometimes those lines bleed a little bit. And so, you know, one of the things we're always challenged with too is like, especially some of the technical shows, is like, okay, how far can we actually go on this? Because we're not, we're not like bringing up a computer screen and like going step by step. So it's always uh, I'm sure, and I'm sure you felt that as well on your own show. It's always a little bit of a balance.

SPEAKER_00 6:46

But I also think like the top-down view conceptually can be more helpful than the click this button. Yeah, there's plenty of click this button manuals and tutorials out there for sure.

SPEAKER_01 6:57

Sometimes it's nice to be able to show something, but at the same time we do video and audio, you don't want to show something, then someone can't see it. And so, any like you said, you don't want it to turn into tutorial hour when you're just trying to discuss a subject intellectually, and it might need to get into the weeds a little bit so everybody understands it, but at the same time, you're not trying to teach this subject in that.

SPEAKER_02 7:20

That's that's right. And I think that the way that we've always posited our show is sort of the idea of these are discussions you'd be having over drinks or coffee or whatever with like your colleagues and friends or over at the you know, eating lunch or something like that. Um, you know, in those situations, if you think about it, rarely is somebody breaking out a laptop and being like, actually, let me show you exactly how to do that in fusion, you know, whatever. Like, you know, these so these are more like kind of conversations with your crew kind of thing.

Coffee Talk Versus Tutorial Mode

Tech Lust And ROI Decisions

SPEAKER_01 7:43

Color and coffee, exactly. Exactly. Yeah. So I have to admit, one of my favorite episodes that you guys ever brought up was the one on technology lust, because it really hit an internal struggle that I have, which is wanting to keep my suite updated, but at the same time having that lust for new technology and being able to find that fine line of do you really need this or do I really want this? Do I just lust for this? And I hadn't really heard that as intricately spoken about as you guys did that. And it really just hit me. I was like, I'm lusting. It was amazing how you guys put it, and it perfectly hit what I think a lot of us feel, which is we want to have the most updated stuff. We want to be able to present our clients with the most accurate and the most efficient setup. But at the same time, there gets to be a point when it's just us wanting to have the latest gear.

SPEAKER_02 8:36

Yeah. Listen, as I said before, one of the things that about our show is that we talk about things that are relevant to us and things that we're going through. And anybody who knows me over the past 20 years, it's like go to you know, go to the doctor and they're like, oh, what do you have? Okay, your cholesterol's good, your blood pressure is good, technology lost check. It's like I live that the good and the bad of it, right? Being on the bleeding edge of things sometimes is difficult, painful. You make a lot of mistakes, and then you're like, why did I just spend all this money on something, whatever? And like I think it's funny, it's great to hear that you like that show because one of the things I've been thinking about is a follow-up about sort of like the progression of like the colorists as we age and like the tech, how that all works. And I feel like recently I've gotten into this vein of just sort of like really seeing the value of things, not just getting it because it's cool, right? And I'll give you a case in point. I think we're gonna talk about this a little later, right? Like that new Stream Deck came out last week or whatever, two weeks ago, right? And like in years past, the first thing I'm doing is popping the credit card information in there, getting it, and it's a race- Oh, it's already on my desk. It's a race to see who gets it first, me or Joey, right? And then we can talk about it or whatever. That was like, I'm like, oh, I'm gonna make a healthy decision here. I already have, you know, whatever, half a dozen stream decks on a couple different desks. Like, no, probably don't need to spend 400 bucks on the new so like it's interesting how that progression too is like you it changes over time, you know?

SPEAKER_01 10:02

Yeah, it's Joey and I were just starting to talk about that offline because I was actually asking for his opinion because I sided with what you were thinking in that technology lust argument internally and said, I actually don't need that. I'm staring at three different Stream Decks right now. And as much as I want it, I don't think I actually need it. And I think I talked to you when that episode first came out about return on investment. For me, a lot of the times it's very easy to see that. But on this one, I think it's more knobs and dials. And they can be more efficient, it can help you work faster. But for me, it's just another thing to program right now. Not that that doesn't make me faster, but it becomes how much time are you spending programming things versus actually working?

SPEAKER_02 10:45

That's well said. And actually, I I will to Joey's credit, even though he's a little bit of an enabler and a pusher on these things as well, right? Like us together sometimes is like the worst combination for stuff like this because he'll be like, I don't know, man, you don't need that. And then like the next day he gets it, and I'm like, well, crap, now I need to go get it, right? But um, I think that's a smart realization that like the ROI, and to be clear, I think there are certain things that can sit in the the realm of screw it, I'm just gonna try it and play with it. Like, you know what I'm saying? Like that Stream Deck in particular is like right on that cusp of like, hmm, I gotta start thinking about this now because this is like four or five hundred bucks or whatever it is. Like, if that came out and it was like 250, I'd be like, ah, whatever, it's just another deal.

SPEAKER_00 11:27

I'll give you an example. I got the new Stream Deck. I'm still playing around with layouts and stuff. I'm using it for my online editing setup. It's great. I love it, worth it. Couple, maybe a month or two before that, Yamaha came out with a Stream Deck collaboration where it's a little audio MIDI panel with transport controls and a motorized fader with a stream deck on top. I'm like, oh, that's awesome. I'm gonna get that. And I bought it, and the top half works exactly like a stream deck, actually works with the stream deck software. It's great. The bottom half is all MIDI and it's not very configurable, and it doesn't really connect to Resolve at all. I haven't found a use for it yet. So that's one of those things where it's like, that looks awesome. I'm gonna buy it. And I got it, and it was pointless.

SPEAKER_02 12:10

It's one of those things too, Jason, where like I think that as the price tag goes up, you just owe it to yourself to really do more analysis, really more research, really understand how things are gonna, but even then, like, I mean, you just discover things. Sometimes things change, or you're thinking like this big, huge VM box that I've been working on as like my main sort of system at home has been a journey. But I will say, even though it's been a journey and some challenging at times, it's also been super educational. Like, even though I've been like mad at it a few times, I also wouldn't trade it for anything because I in the past four or five months I've learned more about virtualization, how pass-through works, grub, Iom U U, like I mean, all these things that make up this thing. And it's just like, no, that's really good knowledge to have. And for us, one of the things testing things out and getting gear sometimes, sometimes they're like test beds for bigger ideas that we're not maybe not quite like so, like the virtualization one's great, right? Let's try virtualization out on a relatively manageable, controllable scale, because we have aspirations in a year or two to get off local systems completely and have a big beast of a machine in the data center. Well, that's a buying into that's gonna be maybe north of a hundred grand. Let's test it out first on a smaller scale, see where the workflow pitfalls are, etc., and see how that all comes together.

SPEAKER_01 13:36

I also think you can over optimize things. I think there is certainly room for you to buy into these projects for fun. There is certainly room for you to, I mean, that's how I feel about my home assistant and my home automation journey. I am not making any money off that, but I am learning a ton and I am adapting that to other areas of my life. But I'm certainly spending a ton of money on it. But for me, it's a, like you called it, a journey. And there are certain things that I feel about being a colorist and a colorist that loves equipment that sometimes you have to choose is this really helping me for work? Okay, is this more of a journey that I'm enjoying for fun? Which I think the Stream Deck fits somewhere in the middle of that. And then there's stuff that just you don't need this at all for any reason, even if it is fun or can help you in your business, it just doesn't fit.

SPEAKER_00 14:28

Sometimes it's fun and cool is reason enough as long as it's not that expensive.

SPEAKER_02 15:05

Yeah, and I would also add that like some of those things feed one another, right? Like, so I think back 10 years ago, this hodgepodge home network with various, you know, unmanaged switches and routers, and I'm having problems with DHCP and all sorts of stuff. And Joey's like, oh man, I just bought into this ecosystem called Ubiquity. You might want to try it out. And the next thing you know, I have 10 devices or whatever sitting in a rack, cameras all over the house. And I'm like, well, crap, now that I got into home assistant and my network's great with Ubiquity, it's time to get into Home Assistant. And like these things kind of like they tend to snowball a little bit. I will say, I don't think I'm ever gonna be in the like collect classic computers just to see if I can run them, kind of like Joey does, but like that's what I was saying earlier that we're we're kind of each other's worst enemy with this kind of stuff. Like, it's like a little bit like, oh, you got that? Well, I'm gonna get that and do it better. Like, so it's a little bit of a competition with it too.

SPEAKER_01 16:00

I think Joey's sort of like that in general. I mean, I slowly started, I had a network that was not working the way I liked it to. And then all of a sudden, like, similar, I have all ubiquity stuff, and now I have five different hotspots in my house. And it really did help a little bit. And that was before I got into home assistant, and having that ability and that confidence in my home network allows me to do other things with my work, with my home office, and things that I would have never been.

SPEAKER_02 16:26

And you start seeing the gaps much clearer, right? You start seeing like once you have a handle on like the big frameworks of things and understand how they play together, you're able to then see where those optimizations happen. But you said something earlier that I think is really interesting because I think it's along those lines that I'm definitely guilty of, Joey's probably guilty of, and many of your listeners are probably too. I look at this two ways. I look at it the as the from field of dreams, build it and they will come kind of thing, where like you can convince yourself, hey, as long as I just get all the stuff, I'll then be all set and it'll be all then the stuff sits in various pieces, not doing what you want it. Like, you know what I'm saying? So, like, there's that trap, but then there's also the trap of like where you overthink the system and you're you become a slave to the system, right? So, like, home assistant's a great example of that. Sure, you can automate your coffee maker to correspond with your spouse's commute skill, like, but then like that one day it's two minutes late and the whole system screws up, or God forbid the home assistant crashed overnight and coffee's not ready at seven o'clock in the morning, like pandemonium, right? I got to the point where like nobody in my house could do anything because they couldn't figure out something it was called, and they push one button and it would set off an automation. So like you can't overthink it too.

Machine Learning Tools That Save Time

SPEAKER_01 17:39

I love what you said about trying to buy everything and sort of building the plane before anything has happened. You have to build everything. One of my business advisors gave me this phrase that actually, when he told it to me, I thought it was crazy. But I actually really enjoy it in terms of larger projects where you build the plane while you're flying. And obviously, it sounds a little crazy if you're talking about an airplane, but I think it's really crucial when you're talking about building a larger business because a lot of us, especially planners like myself, love to over-plan. We'll plan for every circumstance and we'll never get anything done. And I think it's really easy to do that in terms of hardware. So let's jump on to another question that I'm really interested in hearing from you guys. So, recently, uh I want to say like the past couple years, but really especially the last year, machine learning and AI has really not only changed the tools, but changed what our job is. How do you guys feel that machine learning has changed how we work? And how has it personally changed the job for you guys?

Why Generative Video Still Breaks

SPEAKER_00 18:38

Obviously, some of the machine learning-based tools that we get in like a resolve or some of the other tools, things like magic mask, things like the depth map, things like ReLight are really, really useful. The new magic mask version too, especially. Look, it's never gonna give you a VFX quality mat, but in terms of like I have a quick problem to solve in your grade, and you can move on with your life, such a time saver. Little things like I might have done power windows and keys and stuff for a scene. Somebody walks in front of it. I need to hold back that guy that walks in front of it. That's when I'm reaching for these tools. When it's like it doesn't need to be perfect, I don't need to completely rely on it. Softness is my best friend, and you can blur the hell out of it and make it work because it's something distracting out in the ethos of the shot, not the focus. It does such a good job saving time on tasks like that. Then there's also the idea of using the machine learning tools, especially LLMs, as essentially what I like to call the highest level programming language, right? When we started with computers, you were writing essentially machine instructions on punch cards that would tell the machine exactly what to do. Then we moved from punch cards to terminals, and we made things called programming languages that you could put shortcuts essentially, like move this memory to that memory, and then we had assembly, also very hard to write software in. Then we made what was called higher level languages. C, C. That's really all anybody uses anymore, but you get the idea, compiled programming languages. You can actually look at and kind of conceptually see what's going on. If this do that, you know, this variable can have a name, not a memory address. So we're abstracting the instructions of the computer higher level, higher level. Then we get to scripting languages, Python. So much written in Python these days because it's easily human readable, it connects to everything, and you don't have to worry about things like managing memory and how the computer works inside. Well, guess what? One level above the scripting language is the LLM, where if you have a one-off task to tell the computer to do, if you can phrase it clearly in English and describe the needs, both what comes in and what goes out, you can very easily basically you're essentially writing a program for the computer for a one-off task. Whereas I've written programs for automation for years, but it's got to be something repeatable, right, to be worth the time. With the LLM, great example is we had a show where I was doing the censored version for broadcasts of like a major superhero series. They were putting it on cable, they needed to bleep out all the cusses, they needed to blur out the nudity or paint things out or crop things out, whatever. So I would get emails for every episode from the producer with just a written-down list of time codes and descriptions, somewhat haphazard in basically a Word document. Quick little type up on the old chat GPT, take this word doc, format an EDL in this way with the source timecode duplicated the record timecode, et cetera, et cetera. Couple iterations of that, and now I have a quick little script for convert this producer's emails to markers and resolve. Would I have ever written a program to convert that producer's email to markers and resolve before? No, because once that project's done, that's out the window, it's useless, right? But with this, with I can just tell the computer what I want in plain language, now it's like, oh, take this list, convert it to this format, take this EDL, write an email with segment times for my producer. Any task that you're like, oh, this is gonna take me a lot of copying, pasting, and typing, throw that into an LLM and let it handle it, and then idiot check it. Output transcript That's where I love, love, love these tools because they make it very easy to interact with the computer and tell the computer what to do without having to go through a whole bunch of hassle, which makes you do things more efficiently.

SPEAKER_02 22:49

My take on it is that we get a lot of distractions about this AI, ML conversation because it really everybody wants to try to box it into like, it's this, here's the elevator pitch, right? Where in reality, I think that it's very nuanced depending on what you're trying to do. There's different parts of it, there's different sort of zones or areas of AI and ML that do slightly different things. In our industry, we're sort of talking about three major things. We're talking about generative work. I think we're talking about true coding, like writing programs, tools, scripts, et cetera. And then I think we're talking about like analysis, right? Like whether that be workflow analysis or technical analysis, right? I think the first part, the generative stuff, I'm pretty down on, right? And I think that I'm down on it because the world doesn't need to see your cat doing the Macarena, right? Like that's just wasting power, energy, absolutely a gigabyte. Yeah, put pollut polluting the planet. Who the hell cares? With that said, I do see the value of some generative stuff, especially in our industry, right? Like, so if you're gonna be doing stock footage, backgrounds, like those markets come into play, but I'm not looking at it as I mean, it's funny, I have a friend who's producer who's all hot to trot on doing AI ads, and I'm just like, cool, but you can tell that they're AI, like immediately. He's like, No, you can't, nobody can. And I'm like, Yeah, you can, right? So, like the full the buy-in full on to it, I mean, it's getting better, and there's probably some equivalent of Moore's Law to that kind of stuff, but like to me, it's not quite there yet. I value the technicians, the artists who do stuff, but as a tool, as integrating it, like if you're gonna do a project that's all on green screen and the project is lower end and you don't have the budget to do big fancy background or an LED wall or whatever it is, great use of AI. Create an interesting background to put in as a key behind this that I can use for the show. No problems with that, right? The second part of it, the coding thing, I have a lot to say about because this is an area where I've been focusing on recently, because I am really paranoid or jealous about a lot of the young colorists and editors and stuff that I talk to, just sort of get this vibe coding thing inherently, right? Like it's just it's second hand to them to be like, oh yeah, I just wrote an app to do XYZ. Now, 98% of them, as Joey was saying, have no understanding of the underlying technology that's going in there, right? They can't look at the output and go, hmm, see, there's an apostrophe missing here. That's why it's not working. But that doesn't mean that it's bad. And I actually I talked to a person that you guys both know, but she's higher up in the world of post-production and has had some really major jobs. And she was describing, talking about hiring coders recently, and all of them asked her, as like the person doing the hiring, do I have to do any analog coding? Right? Because they were like, as from a job point of view, they were like scared of it, they didn't want to do it anymore.

SPEAKER_00 25:45

See, it's funny because when I hear people even use the word coding instead of saying programming, I just assume they don't know what they're talking about and that they're like 12.

SPEAKER_02 25:53

I understand that might I might have been to be clear, I might have been inserting that word. But the point is that like you know, even even people who are doing this programming are like, why am I gonna like the example she gave was debugging? Instead of having to have somebody get like write debugging scripts that are gonna figure out this code, like just feed that into the LLM and go, hey, look, here's the syntax, here's the language, check this out, whatever. Now, from a personal point of view, I have a quick story about this, Jason, because this was the eye-opening moment for me. So we had a vendor that we were using, like we have a client portal like for upload, right? We were using a vendor that every year when I paid the fees, I was just like, Oh my god, I hate paying this bill, right? And then this year comes around and they're like, hey, we're gonna be raising the prices. And I'm like, Of course you're raising the prices. Like, I already resent you, and now you're gonna raise the prices on me. So literally, like last Thursday or Friday, I was talking to Joey about this, and I'm like, I freaking hate this cost, I don't want to do it anymore. And I was like, hey, Claude, can you make a client upload portal that has the same kind of uploader as frameio does, whatever? And literally within I would say 24 hours, I had something that was totally viable and working, and then spent the next this past week just you know doing like small refinements on it, right? And so, like, this was like out of the box. Like, I'm using Cloudflare R2 storage, which is like S3 storage, uh, HTML port. It's like two files. It's like JavaScript on R2 and an index file on our website, and it's like pretty freaking awesome, all said and done. Like the JavaScript files, like 1600 lines of code, the HTML is probably another 500 lines of code. There's not a chance, zero chance, that I could have done any of that from scratch. And if I was able to do some of it, it would have been mistake laden. Like there would have been mistakes all over it, right? And so, like that kind of thing, I think is great. And I think where this is really gonna excel is that, and I had a friend of mine uh describe this to me one day. He's like, the economy, the creator economy is gonna be really driven by people who can prompt well, right? And I think that's where people like with Joey's skills who understands the analog coding programming world. It's people like him and others, even like you, with like the DCTL stuff that you do with pixel tools, right? Like you have this sort of inherent knowledge of how these tools work. I think you guys are gonna be the superpower ones in this kind of economy because you know how to prompt ask the right questions and even more so look at the answers for how well is this working, right? Now, with that said, I still had a whole bunch load of problems. I mean, the AI hallucinated a bunch of times. I found like one of the biggest challenges with this was as I was iterating, like keeping track of like versions or whatever without using something like GitHub. That was challenging. So, like, I'm like this weekend, part of my thing is I'm trying to transition to like a GitHub site where I can connect Claude to it so like they can kind of talk to each other, but yeah, it's it's cool stuff.

SPEAKER_00 29:25

Now I want to say one thing on the subject of image generation, because this is one where I think a lot of people disagree with me, but they're wrong and I'm right. Whole hog image and video generation. What I mean by that is you type something in and it gives you a video or an image or some combination of the two. That is never going to be good enough for professional work. 100% it will never happen. And the reason I say that, and you know, we see this on the online all the time. Oh, I made this billion-dollar looking commercial for so and so. It looks just like Hollywood. No, it doesn't. It looks like you'd have eight figures, right?

SPEAKER_02 30:06

Yeah, exactly.

SPEAKER_00 30:07

I think people's standards have gone so low because they're looking at everything on a phone screen that they don't realize that if you actually looked at this, it's garbage. And if you look at it more than once, it's garbage. Coke put out their stupid Christmas ad that every truck had 17 different wheel configurations, shot to shot. It made no damn sense. So yeah, at first blush from a phone at distance, you might be like, oh, this looks like something real. Then you look at it critically and you're like, yeah, this is garbage. And the reason why I say it will never actually get to the point of being really viable is because people argue with me all the time. It's getting so much better. Look at what it looked like three years ago. Look at how fast it's getting better. Guess what? Every step forward it takes towards high quality is half as far as the one before. Okay, and this is what you get into is something called exponential decay. Generative AI is on an exponentially decaying path towards the uncanny valley, and it will never cross the uncanny valley. When we're talking specifically about whole hog image generation, I think those things should be in the professional world basically abandoned. And we need to use it for things like manipulation, mat generation, keys, like Robbie said, backgrounds, components, things like tracking. Tracking for VFX work, whatever useful for generative algorithms and getting better and getting better. But whole hog image generation, make me a commercial of this, or here's a start frame and an end frame that I also made with AI, make me a video, even though you're feeding AI and AI and you're getting AI mad cow disease by that. It's never going to be good enough. It's essentially like saying, my car 100 years ago only went 30 miles an hour, now it'll do 200. In the next five years, we're gonna be able to drive at light speed. Nope, there's still fundamental mathematical and physical limits to the world. And mark my words, full image generation will never cross the uncanny valley in a way that is good enough for something beyond just like stupid cat videos.

SPEAKER_01 32:17

We have the same issue with VFX, and I think that's why there's a lot of people that are saying, oh, you know, VFX were better 10 years ago. And it's because they're not used to this decay, where we had these monumental jumps in VFX for a number of years where they just got so much better every single year. And then, like you said, we started seeing this decay where they are getting dramatically better. But what we're seeing is improvements in water physics or improvements in something that is not as easy to grasp. I think the most common one that people pull out is a power to the Caribbean, which looks absolutely fantastic, but now they're looking on improving the smaller things. Like for Avatar, they're working on water physics and photochaps.

SPEAKER_00 33:08

Because they use machine learning and AI tools to do things like how do I integrate my I've got a really good fire simulation. I've got a really good like water and hair simulation. Well, guess what? Now we got a movie with fire, water, and hair and motion capture. How do we merge all of these things together so it's not separate pieces? So all of these physics simulations can really interact in a realistic way. And they used machine learning and AI extensively to do that in an incredibly convincing way. But they weren't saying make avatar, they were saying I'm gonna make a system to integrate all of these different advanced simulations we're building. They were doing components, not trying to do a whole hog generation.

SPEAKER_01 33:50

And that's where I think we're gonna continue to make leaps. Is for instance, one of the things I know they did is they still did the motion capture of the actors, but they built this massive machine learning system to connect it to the rigs that they eventually built up because the way that the avatar body is built is nothing compared to a human. So you can't just map the motions to that. So they had to do this training to essentially remap the human performance onto an avatar so it still looked natural, and they didn't have to have an artist do that frame by frame, and then do it at 60 frames or 48 frames a second in stereoscopic.

SPEAKER_00 34:24

I think the latest avatar looked so good, and the physics were so good, and the motion capture and performance capture was so good. And knowing how much AI that they used in the back end of that pipeline, I'm like, that is the future right there. At its highest level, that kind of technology will trickle down to our work, not make me a commercial, bloop, bloop, bloop, put it on Instagram.

SPEAKER_01 34:51

To essentially sell our point, there's a bunch of people saying that Avatar 3 looks terrible, it looks worse than the other ones. So that's the point is people are waiting for these monumental jumps in technology. When you get to a certain point, it's about the smaller things. Well, I think as I saw, I thought it was.

Hype Cycles And Real AI Limits

SPEAKER_02 35:10

I think I think, I mean, I think there's two sort of parallel paths, but they're kind of diverging a little bit. And I think I don't know how to best explain this, but I think that part of what we're seeing with the generative stuff and what you're talking about, like VFX have gotten worse or whatever, is hey, you know, for the past dozen years or so, every major VFX facility in the world has gone out of business because guess what? Doing it the analog human way is hard, takes a lot of time, takes a lot of people, takes a lot of money, right? The bean counters were like, well, crap, we can get 90% of this look out of running scripts or doing a machine. Let's do that, right? So always the cost factor driving some of that. But then also I think that it is to a large part, we are being artists, especially, like those VFX artists, whoever. There's smart ways of using this tool, like you guys just described with Avatar, but then there's this idea, there's a narrative somewhere that AI can just replace this person and do their job. And I think that's a little bit of a farce right now and probably forever. And I think here's just some really low-level proof of that. Like I have teenage kids, and you would expect that teenage kids would be all into, and to a certain degree, they are. They're into TikToks and Instagrams and whatever, right? But you would think that teenagers would wholly embrace this stuff. My kids, to a large part, themselves as well as their friend groups, are all about that analog life. I mean, like, literally to the point of being like, hey, do you think I could get like an old like walkman on eBay to listen to like tapes? And I'm just like, uh, what? Right? We feel it as that generation that's kind of bridged that gap of growing up analog, moving into computers and digital. I think the kids they can't describe what's going on about it, but it just feels awkward, weird, strange to them. And so the human part of them is naturally grafting on to a lot of these. I don't mean I don't think it's gonna stop it, right? But I'm just saying I think that there is an inherent thing that's happening where just like when we were teenagers or whatever, like people were like, hey, collecting vinyls, cool, right? I think that same thing is gonna be throughout this age of AI, where people like you're gonna go to a facility who's like, oh wow, you have real human colorists. Like, we want to work with you guys because there's humans pushing the buttons. Not like I think that's already happening to a certain degree, and it's there's gonna be some pain in between. But I do see an economy, a media economy, where eventually things will be labeled that they're produced with, I mean, they already see that, right? That produced with AI, that it was made with AI, that it was fully AI. And then I think we're gonna tag the same thing the opposite way. I think we're gonna be like real person, real editors, real colorists, and it's gonna be really interesting to see how that divide splits. My money is on people will generally gravitate towards the imperfections and the charm of human-made stuff, even though these tools exist to do it. And I'm not saying that nobody's gonna watch that stuff and nobody's gonna use that stuff, that's certainly not true, but I do think over time, as this plays out, we will see the human factor step up a little bit more than it is now.

SPEAKER_00 39:08

Can I tell you the age thing is very interesting because I have a friend who just kind of relatively recently moved from a very small team to a large corporate team, and he essentially told me that the large corporate team run by a bunch of like older kind of boomer level age people are really, really optimistic about AI like revolutionizing their business and doing everything. They're like, just do it with AI, it's gonna work with AI, we're gonna AI this, we're gonna AI this that, we're gonna AI that. And like all the 20-year-olds that are just starting in the company are like, you know, that's gonna fall flat on its face and not gonna work, and it doesn't do that, right? And the older guys are like, no, no, AI everything because I saw them on the news.

SPEAKER_02 39:57

That's the point that I'm trying to make much more real about their expectations. That's the point I was trying to make earlier, is that like I think that people don't realize how nuanced and segmented the conversation is, first of all, and then two, there's just this narrative pushing that it's a fix-all for everything, right? And it's just like, no, it's not play it out a few more years. We might be able to to a point where at one point in time, whatever, we have AIs that can control robots. So if you have, God forbid, you have that brain tumor, we can just put you on a table with an AI-controlled robot that does nine out of ten times does a better job than the human neurosurgeon, right? I don't know. Right now, my money is still on the human neurosurgeon to do the work because you know what? If that AI gets to and goes, hallucinates for a second and goes, nah, that part of the brain.

SPEAKER_00 40:47

It's the right side of the brain. Right.

SPEAKER_02 40:49

That brain matter looks like tumor. That brain matter looks like tumor. Let me just take out that brain matter, right? Like, I don't want to be the first person to try that on. You know, so I'm not saying that it can't get there, but I think the next jump is a year, two, three years away is a farce. I think that to get to that point, we're talking technology, just doesn't work that fast, even though we try to make it think like that. Like, what's in the labs right now at these big AI companies doing all sorts of crazy stuff? You and I are not touching that for a long time, right? It's like skunk works in the military, right? Like the SR7, I mean, that's that one thing I always think about too, like military planes and stuff, right? Like weapons or whatnot. Joey knows it's probably better, but what? Like the SR71 was developed in like the early 70s. It didn't even fly until that.

SPEAKER_00 41:35

I think they started late 60s.

SPEAKER_02 41:36

Right. And it didn't even fly publicly, what, until like the mid 80s? Like we didn't even know about it, maybe until the 80s. There's the money factor, there's the control factor. In like, even if we get to that point, right, where we have this awesome artificial general intelligence kind of AGI kind of thing, right? The idea that it's going to be available for everybody to use is just crazy to me. Like, no, the powers that be are going to sequester and keep the best stuff. I just think that people, this narrative is just it's full of what ifs and kind of sticky situations that maybe are not totally true.

SPEAKER_01 42:07

It's a panic response, I think, because obviously everybody's a little worried about the what about me. You can see it in, I think this week or last week, we had a big stock market shift based off Anthropic Who Makes Claude came out with a report, and all of a sudden we see the shop market shift by like five or six percent. And I think it's all just based off a panic response as opposed to actual knowledge of what can this do.

SPEAKER_02 42:31

I mean, I don't think it's totally not true, but I also don't think it's totally true either. So, like, there are absolutely certain roles, responsibilities, and tasks that can probably wholesale get like if your job is at a bank, right? And your job is to analyze transactions for, I don't know, say fraud, right? I can guarantee you the AI can do that work faster, better, more efficiently than a bunch of people sitting at uh, you know, in cubicles can do. But at that same time, I don't want to just take the AI's word for it. There still needs to be somebody to manage that. Like the AI is right now is only like that's the other thing that I think is important in this conversation is that like there's this assumption that AI tools are just they are already artificial general intelligence and that they think for themselves and that they're aware of things when you're not using it. And that's really not true. If you talk to Claude or Chat GPT, it's not thinking about your situation in the background or just pondering it when you're not using it. It's not gonna come up and be like, hey Jason, do you remember a week ago how you were talking about that one coding thing you were trying to do with that DCTL? Well, I thought about it in the intervening two weeks, and here's a good idea for you. Like, it just doesn't work like that right now, right? And so, like the idea that you're gonna have AIs that are taking, and I said right now, I mean, I know people are working on that, but like I think the idea that people are gonna be smothered by AI is a murkier picture than I think the narrative tries to make it right now.

SPEAKER_01 44:07

Murky is probably the best word to describe it. I was using Claude the other day to help diagnose a DCTL error I was getting. And I eventually had to diagnose it myself because we kept going in this infinite loop where it kept suggesting something else, or what about this, or what about that? And when I eventually figured out the error, it was a pretty pedestrian error, and it was easy to fix, but I feel like Claude would have never gotten there in the sense of like it just kept going in other directions, trying more complicated.

SPEAKER_02 44:38

So I I read this interesting bit of research the other day that one of the problems with LL models in general is that it's sort of just like kind of a pull from you know how people say all the time that social media is addictive. It's a little bit of a pull from that, where a lot of times the AI, the models are trying to, they're guessing, they're trying to be predictive and not really knowing. And so, like, that's one of the reasons you get those loops like that. But I would add, and I I've had similar experiences with Claude. One thing that I realized as I was describing building that portal, one of the things that I had another problem with the VM box that I was doing this working on this week, uh, I'm running Proxmox as the virtualization server. That I was like, upload a screenshot, be like, what's wrong? And be like, I don't know, like that button should be there, right? And I realized, I said, What is I asked it, where is your knowledge up to date? It came back and was like, Oh, my knowledge stops at November 2025. I'm like, Cool, well, it's March 2026, and there have been two or three major updates of the software since then, right? And it didn't know anything about that because the model that we're using in public had basically stopped like up to that date. And everything since then, it has to go. So, what I had to do was I had to literally give it. The links to the release notes and the manual for the updated version, and it took 15 minutes to go through that, and then it was much more intelligent about that. But this assumption that, like, it just is omniscient and it knows everything about everything through whatever changes all the time is also a farce. It's all it's doing is it's building neural links, contextual links or whatever, with data that it's trained on.

SPEAKER_00 46:19

And if you don't feed it the right way, the last thing I want to say about AI in general is that whether it's an LLM, whether it's image generation, whether it's whatever, it can only output various mutations of existing work that has been fed into it. It is at its core a gigantic automated plagiarism machine. It cannot do original thinking or original work. Does that make it useless? No. Lots of derivative work is very useful, especially when it comes to things like workflow and tools and stuff like that. But there is no such thing as artificial reasoning. That's yeah.

SPEAKER_01 47:03

And to pull back your example about what you were talking about, Robbie, about is machine learning and artificial intelligence gonna be good at fraud detection? Yes, it might be good at detecting potential fraud detection, but I would still want a human to double check those frauds.

SPEAKER_00 47:20

It's gonna be good at detecting the fraud that we've already detected. And this is an interesting thing about using AI in any kind of security-focused domain, is that again, it can only use what it was trained on. So guess what? You are by nature opening a gigantic hole for someone to come up with a new type of fraud that that AI doesn't know about. And since people are relying on the AI to detect the fraud, guess what? New type of fraud's gonna sail right through for a while until some human figures it out and feeds it into their stupid AI.

SPEAKER_02 47:52

A great model of that, Jason, is you're talking we were talking about the ubiquity stuff earlier. So I got cameras, doorbells, etc. Right. And so ubiquity is bought all in into this AI security model, right? Of detecting presence of people, whatever. And the best one that I've I just had to laugh at was that the other day my son was outside doing something out front playing, and the doorbell detected him and detected him, and it was like, I got a text message, you know, like you're middle of the night, like an intruder attacking your house, right? And this was just like little man near door. And I'm like, little man, right? Like, so like it didn't know child, because he's like five-one or whatever, right? It was like, hmm, he's in between child and adult, so I'm just gonna call him little man, right? It's like to Joey's point, somebody at some point had identified through the training, oh, that's just a little man, right? As opposed to big man. It's just kind of funny that way.

SPEAKER_01 48:45

For everybody that's listening and watching that is intimately concerned about AI taking your job or machine learning taking your job, I think you're gonna be okay for a long time. And I think for the discussions that you have with coworkers and family that are concerned about that, we have a little while until that, if that ever happens, based off what we're seeing in in our community and in tools that we're seeing right now.

SPEAKER_02 49:07

Um Yeah, and I would just asterisk that by Jason by saying that with that said, I think it behooves everybody, just like any other tool set, right? Just like a new plugin, a new app, whatever, a new workflow, it behooves everybody. I think the idea that like this is useless, I'm never gonna use it. I'm not into that camp. But I'm also not into this camp of this is the savior for the universe that's gonna make us all one day sit on the beach drinking Mai Thai's while machines are doing bajillions of things. We are clearly in the middle of that. And I think that for us in the post-production production world, researching these tools, knowing these tools, understanding these tools, even developing some of these tools to a certain degree is great. It's not gonna hurt you and it's what you need to be doing. But the idea that just like it's just the savior, I can't buy into right now.

Suite Design For Speed And Calm

SPEAKER_01 49:55

Just like knowing your post-production tools, knowing the limits of them, you may not need to know every single thing about them, but knowing what you can do with them and knowing that there's resources out there to help you learn the skill if that's something that you discover on a project that you need to do. For instance, you have an XML or an EDL that needs to go into an LLM to learn something more about it or to fix that. You may not need to know that immediately off the bat, but just knowing that that's available for you to jump into if you have the time to is something that you should keep in the back of your mind.

SPEAKER_02 50:24

And I think people need to understand there's a difference between replacement and time efficiency, right? And I think that Joey and I were talking to a client of ours, what, maybe a week or two ago, who was telling us about how they've been experimenting with AI tools to do, like they go out and they're production company, so they shoot a ton, uh, AI tools to do logging for them, right? And it was an interesting conversation because the idea wasn't, oh, we have to replace everybody in our team who's doing logging. The idea was like, I've just taken something that has taken a person a week to do, and I've condensed that down into a day. And now once that middle step has been made in a day, I can now utilize that person to organize further, add some more tags or whatever. So, like something that would used to take a week is now taking two days or three days. That's a great use of AI. They're not replacing that person per se. We've seen this on a lot of levels in a lot of industries, including, I don't know, god forbid, the government, where there's been this reaction to like purge people because of the promise of AI to only relatively quickly bring those people back into play because AWS, how that was working for them. Right, exactly. We bring people back into play because, oh crap, this is not the savior we thought it was, and we still need these people to do, you know, to do some work here.

SPEAKER_01 51:42

Unfortunately, usually something slightly different, but now you may not get the original people that knew the system backwards and forwards and had the knowledge of that system to begin with, which puts you behind the eight ball. So as we start to get towards the end of this episode with you guys, I wanted to ask you guys a question that I know will be near and dear to both your hearts. I know both of you have quite a bit of experience in suite design. And I know you guys have customized both your suites quite a bit. And so I want to ask you guys, what is your favorite feature of the suites that you guys have designed, whether it's your home suite or the suites in the office? Because we color suites are all very personal. I mean, obviously they're supposed to do a specific purpose, they all have to have reference gray. There's some specific things that need to be there, but the rest of it can be very personal. So I'd love to hear what are some of your favorite features that you guys put into your color suites.

SPEAKER_00 52:34

I am at the point now where I feel like with my general desk layout, I am at like maximum efficiency in a good way, in that what I really love is that I've been able to do reference monitor here. My scopes, these are getting really hard to find, but I have a 25-inch ultra-wide LCD that is at an at a nice angle. So basically, if I'm looking straight ahead, I've got panel, then between the panel and the reference monitor, scopes, then reference monitor in one clear view. Then I've got all my Stream Deck XLs and regular Stream Decks all programmed next to that panel in 3D printed housings to be exactly the right locations. And now I spent a good amount of time earlier this year going back and redoing all of my different node tree, color management, layout presets that I use daily to one unify them because I had a couple different like generations of things I was using across different projects, and two really integrate them into the macros that I've already built for this layout. So combining both physical layout, hardware, software, and now actually just sitting down and making like one hero preset to rule them all that can work with any workflow that I do, but everything's in the same place and everything's macroed now, so I can just like press one button to go to the node that I always know is gonna be there, and my UI is always gonna be there. That has saved me so basically starting at the beginning of the year, taking everything I've hodgepodge built over the years and kind of gutting it and rethinking it and putting it all back together in a very similar way, but in a much more optimized and integrated way. That once every couple years, just saying, you know what? No, I'm gonna rewire everything, I'm gonna clean everything up, I'm gonna organize all my software, my macros, my layouts, I'm gonna sit down and and knock out all of the little, oh, I gotta click on this every time. Inefficiencies, both physically and software, has been wonderful. But we talked about this a little bit earlier offline. The number one favorite thing in the world for me right now is that now that I have 120 hertz refresh on my UI monitor and my little cursor moves so smoothly when I'm using my tablet, that just makes me so happy every time that I never want to go back to a 60 hertz UI ever again. And I know it's a tiny little thing, it shouldn't matter. Like we were talking about with gearlust and like, is it worth it? Yeah, that was worth it to me. Every time I use the stupid computer, it's so smooth it makes me happy.

Wrap Up And How To Connect

SPEAKER_02 55:21

The biggest thing about any sweep design for me is the number one factor in anything is cable management and the hidden cables. Like I have like this just belief that cables are just like they give me anxiety. I hate it, right? So I spend a considerable amount of time in my life optimizing my view of cables so that I can't see any cables. To be honest with you, looking around, like I've done a pretty good job. Like there's one or two cables that I can see, but I can't get rid of them. If they could be wireless, I take that, right? But no, I think that the biggest thing that I've done probably over the past year, at least in my home suite here, is unifying everything to the displa same display technology type, right? Including my UI monitors. We actually just did an episode on this where we talked a little bit about the idea of single monitors and then dealing with multiple monitors and how to deal with them and tamaras to do perceptual matching. And if your audience wants to check that out, my general gist of my take was I eventually said, screw it all. I'm just gonna have one big reference monitor and not look at anything else. And I've taken that to the extreme with computer monitors, right? So now everything in my room is QDO LED, like literally everything. I guess the only screens that I have that are not QDO LED are the screens on the Stream Decks, but like that's I can't do anything about that. And that's actually been a really big boon to kind of my feel of like consistency of image, like where I'm like, okay, yep, yep, yep, yep. And it's also like in a small way, allowed me to do things like have more confidence when I'm doing some compositing work or whatever, where I'm not necessarily looking at the reference monitor all the time. So that was number that's probably the second biggest thing. And then the third biggest thing that I recently that I love about my room is just how I've been able to just get rid of a lot of stuff. The virtualization part, I mean, that has allowed me to get from I don't know, probably four or five different machines in a rack down to really just like a workstation and the server. That's really gotten things quieter. It's saved me on a lot of energy. That's been awesome. And then the third thing I would say that my recent favorite thing, and Joey's just gonna laugh at this because he thinks I'm insane, is that I have moved my network storage, my NAS, completely off of spinning discs. Um, everything I have now is all NVMe. An NVMe coupled with 25, I'll say with an asterisk, 100 gig networking. I have a 100 gig card, but don't have anything on the client side that can support that. So I said later on for that. So, but everything's 25 gig now, and you're like, ah, what's the big big difference? 25 gigs pretty cool when you're busting through, taking advantage, make it makes you feel like those SSDs, like those NVMe's, were a little bit more worth it when you're popping, you know, 25, 2600 megabytes of through them a second kind of thing.

SPEAKER_01 58:01

Those are some great upgrades. And I love that everybody has their own thing that they want to focus on. You say Joey thinks you're crazy, but that's what I love about it. Everybody wants their suite and what they focus on their own way. Like there's little things we have to have a reference monitor that's calibrated. We have to have most of the time a control service, but the rest of it is how you want your suite to feel.

SPEAKER_02 58:22

Yeah, to flip it back to the top of the conversation. I mean, some of this for me is always about gear lust, right? Playing back ProRes files, do I need more than 10 gig? No, there's not a logical argument that says I need more than 10 gig to play back, you know, ProRes files for podcasts. But does it make me feel awesome? 100%. This is like, I know because they're a sponsor of your show as well, right? I've driven over the years, Brahm from FSI, I've driven absolutely batshit crazy because I'll call and be like, my delta E is like 0.01 off of whatever. He's like, dude, it's fine. And I'm like, really? You know, so like they're uh people have their own obsessions for me. Speed and accuracy is probably where I get into the weeds for sure.

SPEAKER_01 59:04

Awesome. Well, thank you guys both for coming on the show. It was such a pleasure to have both of you. I really, really enjoyed our conversations. If you haven't already checked out the Offset podcast, I highly, highly recommend you check it out. There is 50 episodes that are jam-packed with super fun conversations if you enjoyed what we talked about on this episode. And for this episode of Color and Coffee, I want to thank our guests, Joey and Robbie. Thank you for joining us.

SPEAKER_02 59:32

Yeah, Jason, thanks so much, man. Always a pleasure to talk to you. And uh thanks for having us on.

SPEAKER_00 59:36

Truly such a pleasure. Be back on here one more time. I'm sure we'll we'll do it again. So thanks, everyone.

SPEAKER_01 59:43

Really such a pleasure. I can't wait to have both of you on the show again. And for this episode of Color and Coffee, we'll see you guys for the next one. Happy grading. And that's a wrap. Be sure to follow us on Instagram, YouTube, and your podcast app of choice. Search for at Color Coffee or at Color Coffee Podcast and join the conversation. If you're using Spotify or Apple Podcasts, please leave a review. Huge thanks to FSI, Demystify Color, and Pixel Tools for sponsoring the show. Until the next episode.