TRANSKRIP 🔗
WEBVTT
1
00:00:00.080 --> 00:00:12.300
213 locations across 18 countries, and we shot it in 21 days, one location This is Ryan Kavanaugh. He's the producer of Doug Liman's upcoming film, Bitcoin, a $70 million AI hybrid production.
2
00:00:12.480 --> 00:00:22.700
In this episode, we get into where that $70 million budget actually went. The beauty for the actors in our case is we paid them full freight. Director got full freight. Department heads got full freight.
3
00:00:22.880 --> 00:00:34.400
We have 100 AI artists now that work for us, and we had over 130 day players. That's more than Doug has ever had on any movie. But what I keep trying to tell people is the savings isn't in people. From VP Land
4
00:00:35.720 --> 00:00:52.040
at AI on the Lot. [upbeat music] This is Inside the AI Studio. Ryan, tell me about Bitcoin. [laughs] Well, it's this thing that people buy. No, I'm just joking.
5
00:00:52.070 --> 00:01:03.040
[laughs] So like, about, I don't know, three years ago, after I'd sworn off the movie business forever, um, I came across this really interesting story through actually an AI business that I owned that wasn't in the movie business.
6
00:01:03.660 --> 00:01:11.400
It was a AI that does AI for a lot of the social media and, um, the story was fascinating, so I was like, "I gotta make this story." Developed it on my own.
7
00:01:11.460 --> 00:01:20.860
Hired probably the best writer I've ever w- worked with, Nick Schenk, and it's, uh, the true story of a guy who may or may not be Satoshi. His story in general was just fascinating.
8
00:01:20.920 --> 00:01:35.300
Like, whether he was Satoshi or not, what he stood for being, you know, basically against the corruption and greed that had taken over Bitcoin and for the original Satoshi vision, and so developed it, got a great script, got Doug Liman attached after some time.
9
00:01:35.900 --> 00:01:42.840
Great cast. Got my first budget back. We're independently financing $150 million. Then I was like, "Oof." I definitely didn't wanna go to studio with it.
10
00:01:43.040 --> 00:01:49.300
Wanted to keep control of it and realized that was not happening at 150. So went back to some of my AI contacts.
11
00:01:49.420 --> 00:01:54.190
I'd been reading a lot about deals with Runway and Midjourney and so on and so forth, and went and met with...
12
00:01:54.900 --> 00:02:15.380
You name it, we met with them, and we left every meeting with the same outcome, which was they are never gonna get it, at least not in the short term, and it's not because of tech, and that's where people get really, really confused in this business is they all have tech that could theoretically work, but it's that people look at it as pure tech, and AI is not pure tech.
13
00:02:15.500 --> 00:02:21.180
AI is, you know, kind of 50%, I'll call it tech or generation, but it's 50% art.
14
00:02:21.400 --> 00:02:43.740
And so when all these studios are making deals where they're bringing in chiefs that don't have backgrounds in physical production, I mean hands-on production, like on set, having sat through what happens when this camera breaks, what happens when the light won't work this way, what happens when you can't shoot from the corner, you know, what happens in post-production if you have something that's, you know, overlit or underlit, what happens if an actor gets sick and the schedule leans off and, you know.
15
00:02:44.220 --> 00:03:01.960
So all these things, y- y- you had studios and production companies going to tech companies that were AI saying, "Help me make this movie," and we just saw the writing on the wall that that was never gonna work because, sure, they'll be able to get to some first frames and get a video out, but it's gonna look wrong.
16
00:03:02.380 --> 00:03:09.320
Mm. And so we went down the path of starting to hire our own AI people and, you know, figuring out, can we make this movie cheaper using AI?
17
00:03:09.780 --> 00:03:16.400
We ended up hiring, and we put together our, our own workflow and our own team, and we have a hundred artists now that work for us. AI artists, what we call them.
18
00:03:16.920 --> 00:03:27.880
Set up our own studio in London, which we, we bought and built out, um, and then it's a lot of trial and error. So it started with seven months of pre-production with a 50-person creative department because
19
00:03:28.920 --> 00:03:43.340
I don't care who says differently, I don't believe any time in the near term you're gonna be able to make an AI movie that looks good without having every first frame, every location build down to spec, down to, you know, architectural perfection.
20
00:03:43.580 --> 00:03:54.020
So we had to spend $15 million of our budget to do that, and then we shot the first movie, Bitcoin, in all AI, and it's what we call performance capture.
21
00:03:54.120 --> 00:04:01.760
So for us, what's really important coming from the industry is that, you know, it's, it's artist and creative first. So a- all the actors are real.
22
00:04:01.769 --> 00:04:11.600
We have our own form of performance capture, as we call it, on our proprietary gray stage. The director directs them just like you would. The DP shoots them and, you know, it's a normal kinda crew.
23
00:04:11.700 --> 00:04:27.060
It's just that when you're shooting, the entire focus is just on the actor. So in our case, we had a 72-day physical shoot with 213 locations across, like, 18 countries, and we shot it in 21 days, one location.
24
00:04:27.120 --> 00:04:33.060
Not, not one plus one, just one. Our, our gray stage. Turnaround's five to 20 minutes.
25
00:04:33.320 --> 00:04:46.560
We shot the m- the leads out in 19 days, and we were able to drop the budget below the line from one, call it one twenty, one thirty, down to forty, which now would be twenty-five because 15 was spent on pre-production.
26
00:04:46.640 --> 00:05:00.240
And what we did is we used the movie to build our own AI system that integrates with our AI
Load full transcript
artists and protocols in shooting and with the gray stage to do all that pre-production stuff automatically and save that $15 million.
27
00:05:00.300 --> 00:05:08.820
So here we are. We f- we wrapped a few weeks ago. Um, one of the beauties of AI is you're able to see scenes right away, and we're really excited about the movie.
28
00:05:09.280 --> 00:05:15.420
You've answered a lot of my follow-up questions [laughs] in that answer. But, uh, yeah, 'cause I was wondering, you know, we heard about the $70 million budget.
29
00:05:15.620 --> 00:05:20.469
You know, if you were producing this today, you sort of answered that, like, if you started today, I mean, how... If we started today- 'Cause this was over, like, a two-year process...
30
00:05:20.500 --> 00:05:30.980
it would be 20 to 25 million below the line. Mm-hmm. Plus whatever, you know. I mean, the beauty for the actors in our case is we paid them full freight. Mm. Right? Director got full freight.
31
00:05:31.240 --> 00:05:36.980
Department heads got full freight. Ironically, so we had over 130 day players. Mm.
32
00:05:37.340 --> 00:05:46.100
That's more than Doug has ever had on any movie, more than Born, more than Day After Tomorrow, more than, uh, or Live Another Day, more than, uh, Mr. & Mrs. Smith.
33
00:05:46.600 --> 00:05:55.080
And so the difference is that we're able to bring them in and shoot all of them, all 130, in less than two days, and we have a, a proprietary capture system.
34
00:05:55.820 --> 00:06:05.883
And, uh, you know, so yeah, today it would be 25 million plus whatever's above the line and- We never really expected to be a production services business either. Mm-hmm. We were just doing this for our own.
35
00:06:06.404 --> 00:06:18.084
And as people came through, studios, producers, they started asking can we do their movie. And so we have over a dozen projects now in various stages, and it's really cool to see,
36
00:06:19.144 --> 00:06:27.484
you know, like the way we, we kinda look at it is the movie we made, the kind of original IP unproven is like they just don't get made anymore, especially in the mid-budget range.
37
00:06:28.164 --> 00:06:37.104
And there's a lot of very talented people that have stories to tell, and that's what Hollywood was built on. You know, not sequel, prequel, tent-pole, rinse, repeat, right?
38
00:06:37.724 --> 00:06:47.924
And so we're watching people who've had projects for three years that are well-known filmmakers or talent that are like, "The studio said they'd do it for 30, but I can't figure how to get it down less than 70."
39
00:06:48.084 --> 00:06:59.104
And we're able to do that for them, and we're able to show them that it's 100% photoreal and that they get more creative control and less time involved. And, um, that's really neat to see. Yeah, that's an exciting time.
40
00:06:59.984 --> 00:07:09.014
Can you walk me through just high level like what... So what was the production process? You're, you're filming the actors on the gray box. Is it a 3D build? Like what does Doug see when he's looking in the- So it's-...
41
00:07:09.014 --> 00:07:11.024
the monitor... very interesting how sparse it is.
42
00:07:11.184 --> 00:07:22.884
For us it was, it's, you know, call it, think of it as a just a big studio that is gray, sounds like it sounds, that has all kinds of different cameras, a very specific set of lighting that doesn't change- Mm-hmm...
43
00:07:22.934 --> 00:07:32.864
it's lighting for the AI. That has blocking in gray and a lot of different shapes around the walls. Again, things that we've perfected to make sure that it captures right.
44
00:07:33.044 --> 00:07:44.104
What the director does is i- our system, as I said, you input and you kinda get out all the location builds. So before you start any shot, the location is blocked.
45
00:07:44.194 --> 00:07:54.034
And what I mean by blocked is we have a proprietary AR system that you put over the stage and you can literally see photoreal. Mm. Like this is what it's gonna look like. This is where this table is. Mm-hmm.
46
00:07:54.044 --> 00:07:58.424
This is where whatever wall is back there. Mm-hmm. This is- But none of this exists. This all- But none of it exists.
47
00:07:58.484 --> 00:08:04.664
What, if we were shooting ourselves or they were shooting us, it would be you and I having a conversation- Mm-hmm... sitting in two chairs. Yeah. Likely not the chairs that would show up.
48
00:08:05.184 --> 00:08:14.784
The rest would be blocked with gray and, and everybody's able to use the VR projector to see where things would be. But that way now the director just directs normal.
49
00:08:15.044 --> 00:08:23.904
Normal cameras if he wants to use his close-up, if he wants to use wide, if he wants... Whatever shots he wants. You're tracking, moving because he's capturing the actors in gray.
50
00:08:24.244 --> 00:08:33.324
Our job is to then take that kind of what we've beamed down, right, the location, along with the first frame and build the world around them.
51
00:08:33.724 --> 00:08:47.614
So, you know, the director still has to capture the action, still has to capture the interaction, and what's interesting is the actors when we ask them, you know, how, how they liked it, first of all they loved that they shot out in 19 days.
52
00:08:47.644 --> 00:08:55.563
The setups are between five and 20 minutes. The longest- Mm... setup was 20 minutes. You know? So- It's all lit. It... Well, it's lit. You're not building anything. Mm-hmm.
53
00:08:55.644 --> 00:09:04.644
So what the A- our AI program does, amongst many other things, is that it also tells you exactly what props for that scene
54
00:09:05.564 --> 00:09:12.184
they, we should use because there's certain things you don't wanna have to recreate in the AI, right? Mm. So like you're holding an iPad.
55
00:09:12.724 --> 00:09:20.304
It would tell me, yes, you need, he should hold the iPad 'cause that's gonna be harder for us- Mm-hmm... or it might be a, a
either very expensive or very tricky or it might not look right.
56
00:09:20.864 --> 00:09:31.164
And so you've got your AD, you've got your DPs, you know, I think we had 80 crew members. But what I keep trying to tell people is the savings isn't in people, it's in time.
57
00:09:31.813 --> 00:09:42.664
The department heads are more important than ever, r- right? 'Cause not only are, are, are in pre-production they crafting really the whole movie. So for us it's very much like an animated workflow.
58
00:09:42.744 --> 00:09:48.384
You watch the whole movie before you shoot it, and you can change scenes, you can, you know, manipulate, you can modify before you shoot.
59
00:09:48.844 --> 00:09:58.964
But then in post, they're working through our system which then they get to communicate with our team through the VFX sup. And, um, they're able to...
60
00:09:59.004 --> 00:10:08.564
You have your lighting, you have your camera, you have your set design, you have, you know, your cinematographer all coordinating and communicating through the program of what needs to be done.
61
00:10:08.624 --> 00:10:18.844
Because our AI artists are all trained in production and work in production now, right? And I think we're the only ones so far to do that, but the vision's the director's vision.
62
00:10:19.044 --> 00:10:26.174
And so he has his team of department heads that have to make sure that, you know, our team is actually building the vision. Just like,
63
00:10:27.204 --> 00:10:34.604
you know, if, if the director was shooting a real, you know, like call it we're doing a concert- Right... and he has a vision of the concert.
64
00:10:35.104 --> 00:10:40.784
Well, while shooting he would've created a vision using all those guys, the cinematographer and the DP and the lighting.
65
00:10:41.184 --> 00:10:47.244
Well, now you've shot it and you kind of, you know what the location is, but you gotta create the look and feel after the fact.
66
00:10:47.284 --> 00:10:56.704
And so they become more important in the post process, more important in the pre-process, and actually it's the shooting process that they're less involved. Um, and so yeah.
67
00:10:57.424 --> 00:11:07.104
Was there a version, like was there a full pre-vis version made of the film before anything was shot? There was quite a bit of it. So as I said, we built this program- Mm... while doing that.
68
00:11:07.644 --> 00:11:18.724
We had every s- storyboard, every first frame, every location, but we didn't actually have the animated pre-vis, which now, again, the program does and, you know, could do...
69
00:11:18.744 --> 00:11:29.304
Without, if you're not putting too much into it in terms of like vision, right, you're just wanna put the output, it does in days. And, you know, we have people that work on it to make sure that it's done right.
70
00:11:29.784 --> 00:11:38.034
But it not only will do the animated pre-vis, it actually will generate the whole movie for you to watch. Again, it's not gonna be in a quality you would put out. Mm.
71
00:11:38.064 --> 00:11:51.644
We're not trying to replace actors, but the director can watch the scene in photoreal quality and be like, "Oh, I kinda didn't realize that's how it would come out," or, "I want the actor to act it this way," or even, you know, and we've watched a lot of people change things based on that.
72
00:11:52.344 --> 00:12:00.584
Yeah. And, um, a- and also like would help just before you shoot anything you already know what you wanna get, so you're not wasting time shooting- Yeah, so-... stuff that gets cut...
73
00:12:00.744 --> 00:12:10.924
we call it the 100% certainty you never had before. 'Cause when I ran a studio I literally used to say, "Why can't we go hire, you know, USC film students to animate the whole movie," right?
74
00:12:11.414 --> 00:12:21.374
For inexpensive, so I can watch the whole movie, and we can test it before we shoot it. Like, instead, you know, the world today is you go spend $100 million making a movie, right?
75
00:12:21.494 --> 00:12:27.974
And it's usually a 50, 60, 70-year-old, you know, green-lighting it, 50, 60, 70-year-old making it, right? Making all the decisions.
76
00:12:28.634 --> 00:12:34.774
And you finish, and then you go put it in a theater after you're finished, and you show a bunch of 18-year-olds and say, "What do you think?" And they're like, "Oh, it sucks."
77
00:12:34.894 --> 00:12:44.454
Now you go spend $2 million trying to fix it, right? For the first time ever, you actually can make it, and again, you're not saying this is a quality I would wanna show.
78
00:12:44.474 --> 00:12:51.594
You're not saying this is something I'm gonna go sell tickets for. But make it where you can see at least does it work as a concept. Do I like that scene? Do I not like that scene?
79
00:12:52.094 --> 00:13:02.054
And i- if you so want, even test off that before you go spend a dollar. But you think that's also gonna open up more validation pipelines or content creation pipelines?
80
00:13:02.674 --> 00:13:07.454
Someone does something that's popular on YouTube, easy to validate. Yeah. Turn it into a film.
81
00:13:07.894 --> 00:13:17.844
I think what it is is it's more, you know, as we're getting all these kind of work for hire projects, what we're, we're seeing is the studios want a lot more certainty up front. So they want proof points, right?
82
00:13:18.474 --> 00:13:25.614
And the proof points being the fact that we can allow them to kind of see pieces of it, see i- if they can't undertake.
83
00:13:25.654 --> 00:13:37.154
We have this one movie where there's-- it takes place in, in Formula 1, and they're concerned that, well, the budget's pretty small, and F1 just came out. And so it's not about F1, but it takes place in F1.
84
00:13:37.514 --> 00:13:47.994
But even so, having-- people are so used to this big spectacle, epic F1, and if we have this teeny budget, like, it's gonna be a disappointment. So we were able to show them, "No,
no, no.
85
00:13:48.214 --> 00:13:58.894
This is exactly what it would look like." And they're able to say, "Oh, wow. Okay. Go." You know? Yeah. Or i- in certain cases, showing them an entire previous movie, and
86
00:14:00.274 --> 00:14:05.444
we've seen a lot of people come back, rewrite, change, you know, and be like, "Oh, well, it's green lit subject to the following."
87
00:14:05.654 --> 00:14:14.284
And so I think more of that, kind of the YouTube building out proof concept, I, I know a lot of people are working on that. That's just not what, you know, kind of our- Mm-hmm... forte.
88
00:14:14.394 --> 00:14:26.294
Like ours is we wanna make our two or three movies a year. We're doing a lot of work for hire just because we're excited to help the industry, and, you know, it makes good money. So [laughs] Yeah. New pipeline. Yeah.
89
00:14:26.344 --> 00:14:38.354
Get more, get more stuff made. R&D time and stuff out of the picture, how did the kind of flip of time for pre, I know production's shorter, but then and post in this hybrid workflow?
90
00:14:38.374 --> 00:14:43.814
Well, so, you know, as I said originally, the, the seven to eight months- Mm... was absolutely necessary.
91
00:14:43.854 --> 00:14:55.024
And I, I do believe there's gonna be other AI movies coming out that are not gonna look good, not because they didn't have the right artists or the right tools, but because they didn't prep it right. Mm-hmm.
92
00:14:55.054 --> 00:15:04.734
They didn't have everything that was needed because it is very expensive. We took that and looked at it as an R&D for o- building a program that we now have 60 patents on.
93
00:15:04.914 --> 00:15:07.594
So it gives you any idea how many different things it does.
94
00:15:07.694 --> 00:15:25.034
That, to me, now we've taken pre-production from, you know, seven months, and, you know, as studios want more certainty on a big movie, $15 million, $20 million in pre-production to do testing and boards and hold to, you know, to minimize whatever the cost of running, uh, you know, some tokens are.
95
00:15:25.514 --> 00:15:35.214
And then it could be 30 days or less for pre-production. Like, we have a project we got yesterday that literally is going in July, and we-- we're, we're just doing the pre-production. Like, you got a script yesterday.
96
00:15:35.534 --> 00:15:41.014
Yesterday. And now it- Yeah... pre-production's July. Actually, the script isn't even locked. Like, that's how... But because we can do that now. Right.
97
00:15:41.094 --> 00:15:46.254
Um- I mean, do you feel like this has also opened up this more of a kinda loop? Instead we have the traditional pre-production post.
98
00:15:46.354 --> 00:15:55.174
Do you feel like it's kind of turning more into a- Well, like in that case, what it's really... It's just, it's just tightening timelines. Like I said, what everybody needs to realize is AI is about time, not people.
99
00:15:55.654 --> 00:16:03.514
And so in that case, you run it, you have
your storyboard studio, you have your first frame studio, you have your location studio.
100
00:16:04.094 --> 00:16:10.194
You just put all the department heads in right away saying, "Go work with your director and, you know, get, make these story..."
101
00:16:10.234 --> 00:16:15.314
It's, it's all natural language, so change the storyboards to, you know, whatever is gonna make sure it's his vision. In some cases,
102
00:16:16.274 --> 00:16:28.394
they, they like it as is because they're shooting it, and they can change it after the fact, so the location, as long as it's right from a spec perspective, if they're like, "Well, I wanna change the wall to a different color" whatever, they know they can do that.
103
00:16:28.834 --> 00:16:32.913
So sometimes they don't even do anything with this, but they can get as detailed as they want.
104
00:16:32.994 --> 00:16:44.894
So f- 30 days is huge amount of time to get in all day and have department heads, and they all, it all coordinates together, like making changes to looks and feels. It actually has every single shot is pre-planned.
105
00:16:45.134 --> 00:16:55.774
It does all the scheduling, all the sides, and what it does is it coordinates within both actor schedules, crew, but also the AI to make sure it's the most efficient way to shoot.
106
00:16:55.814 --> 00:17:06.154
And so if something changes, it automatically also re-triggers. So your system's also kinda handling the production of scheduling, the kind of- All... breakdown process as a central. Yep. It does all of that.
107
00:17:06.314 --> 00:17:16.724
And, um, as well as literally it's telling you the shots. So the director doesn't have to use that, but it does give you shot by shot. It also gives you the VFX breakdown. Mm.
108
00:17:16.724 --> 00:17:27.454
Meaning the VFX is an AI, or we call it VFX 'cause it's a VFX pipeline. Mm-hmm. So the pre-production is really sped up. The production is really sped up. We drop production time by 70%.
109
00:17:28.094 --> 00:17:38.014
Um, and then post is probably the same, if not a little bit less, but the difference is that in, because of the way AI works, whereas in other movies, take Avatar, right?
110
00:17:38.114 --> 00:17:48.534
Like, not to compare Bitcoin to Avatar, but there are a lot of similarities in how it was shot, right? Mm-hmm. Avatar's all mocap and green screen and, you know, and we're all gray s- uh, stage.
111
00:17:48.634 --> 00:17:59.594
And so we're at 270 locations. They're in these crazy worlds. Yeah. So, you know, Jim Cameron has to make that, and before he can really see any of it, he has to have the compilation put together.
112
00:17:59.634 --> 00:18:06.154
He has to have the, the, the entire background made. He has to have it go through a post-rendering, and that's a year probably, right?
113
00:18:06.874 --> 00:18:17.674
For us, we just showed 10 minutes of it at Cannes, not totally complete, but good enough for, you know, government work. Um, and we were able to do that two weeks after shooting before we even had a cut of the movie.
114
00:18:18.234 --> 00:18:27.534
And that's also part of what speeds it up is you can watch things and change them without having a huge expense. You can watch the scene and say, "You know, I want that, that... I don't love the way that the city looks.
115
00:18:27.654 --> 00:18:35.054
I, I, I wanna change the lighting. I wanna..." You can literally modify anything, which is good and bad. You can modify it in production, or you can modify it in post. Yeah.
116
00:18:35.394 --> 00:18:43.434
But you have to at some point be like, [laughs] "We're done. We're locked." Yeah. Well, like Doug came in one day, and he had gone to this place in London called Sexy Fish.
117
00:18:43.774 --> 00:18:51.414
It's a kind of very popular sushi place, and it's very, uh, lively. It's more like a nightclub, right? Yeah. And the night before, he'd gone to Cirque du Soleil.
118
00:18:51.734 --> 00:19:00.534
So he came in, and he's like, "Okay, I'm changing Nobu," so we're shooting a, a sequence- Mm... that was supposed to be Nobu, "for my version of Sexy Fish he calls Koi Koi."
119
00:19:01.134 --> 00:19:04.834
That is a very upscale version of Sexy Fish with Cirque du Soleil in the middle.
120
00:19:05.534 --> 00:19:15.274
So we had to, like, plug it in, run everything, redo it, but we could do that, and within hours, without having to go find set designs and prop.
121
00:19:15.394 --> 00:19:24.934
You know, we were able to do that, and that would've probably been a $5 to $7 million shoot- Mm... to do practically. And instead, for us, it was the same as two people sitting at Nobu.
122
00:19:25.154 --> 00:19:27.214
So you do have to have-- I mean, he did that. Another one which
123
00:19:28.114 --> 00:19:43.364
we got, one of our leads was supposed to have a call from his desk at Antigua, and instead he's like, "I want him on an expedition in Antarctica, in, in Antarctica, in the middle of a snowstorm using a satellite phone while there's, like, wolves over here and this happening," just because he could, right?
124
00:19:43.964 --> 00:19:52.934
[laughs] And so those are the things. And they work for the-- But you also have to have a director who can keep it under control because they could literally spend all day spinning out of control of what they wanna do.
125
00:19:52.974 --> 00:19:55.594
Yeah. Ryan, thanks so much. Appreciate it.