The Book Of Boba Fett
- Thread starter Ar-Pharazon
- Start date
Agreed. It worked well, Ehrenreich did great, especially for the story told.
Nah.
Deep fake technology is getting so good at both video and voice that in a few short years, you could do a ST set right after ROTJ using motion capture and deep fakes- not the stunted, limited deep fake work we got in Mando S2 and the BOBF. I think this is a big part of what the current strike is all about.
That said, I was okay with Ehrenreich as Han, and I thought Glover was one of the high points of the movie.
That said, I was okay with Ehrenreich as Han, and I thought Glover was one of the high points of the movie.
I don't understand why anyone would want fiction that's just churned out by a mindless algorithm rather than representing the imagination, hard work, and skill of talented people. I mean, would a sports fan be on the edge of their seat watching a computer simulation of a tennis match or gymnastics competition? Would anyone be thrilled or nervous watching a computer-animated acrobat on a virtual tightrope? It's the exceptional talents and achievements of people that we celebrate. There's no meaning to it if there aren't people doing it.
I can. It's not about the talent. Every younger generation glued to a phone is just interested in the entertainment aspect of media. There's not much thought to what you just said(imagination, hard work, skill, talented people, etc...). This piece of media occupies a certain space now for entertainment. After it's gone it's unto the next thing to occupy my attention span until the next thing comes up.
To be clear, I think it's terrible but it is already here to a certain degree.
Side Note: Donald Glover really doesn't look much like Billy Dee Williams at all, but you rarely hear people arguing that he wasn't (and still is!) perfect for the role!
This is a fantastic point.
Fair enough but Christopher's post that I was responding to was a lot broader than just deep faking characters.
A few things to unpack here. First off: "Could"? Sure. "Should"? Almost certainly not.
Secondly; as much as I have enjoyed having younger Luke show up, it's only Mark Hamill's direct participation in those scenes that makes it an acceptable use of the tech. Just slapping his likeness that that of the other original actors on a bunch of doubles and calling it a day is not something I want to see, and would set a dangerous precedent. If they want to use those characters in that era more extensively (especially Leia); then they should recast.
Thirdly; it should be noted that his appearances on those shows have out of necessity been very limited. Those shots were crafted very carefully around the limitations of the technology that may seem like anything is possible, but it really isn't. If you tried to apply this on a larger scale AND with multiple characters in all kinds of lighting conditions, it would fall apart at the seams FAST. Indeed, in preparation for that episode they tested the method to destruction to see exactly how far it can be pushed, and how far it can't. So from a practical standpoint, it would be very untenable.
And finally: the process used for the season 2 appearance wasn't a Deep Fake. Deep Fake is a very specific methodology that involves machine learning to build images a pixel at a time from leaned patterns, hence the "deep" from "deep learning".
What they did for 'The Rescue' was a proprietary method developed inhouse by ILM that used a Deep Fake pass as reference footage only, since the resolution was way too low to be usable in the final element (what might look good on youtube, would look like dogshite if you tried it at 4K.) Said elements was produced with basically the same tech they used for Tarkin & Leia in Rogue One, only on a much smaller budget & timeframe. That method just straight up replaces the double's head with a rigged mesh just like any other animated CG character, albeit at a much higher level of fidelity.
What they did with BoBF seems to have been a hybrid of the two methods, with a much more refined high resolution Deep Fake pass done over the CG head replacement; that later of which doing most of the heavy lifting and the Deep Fake mostly functioning to bridge that last little leg of the uncanny valley.
It'd be insane to try and do all of that for multiple lead characters across multiple episodes of TV. Indefensible on moral grounds, and untenable on practical grounds.
Last edited:
I thought that was clear in my post, but maybe not. You are correct.
I tend to agree. But not everyone will. I'm speaking to what will be possible in a few short years, not what studios should necessarily do.
Again, what is not achievable today will be done on home computers years in the future. Look at the fx work being done now, for fan films, vs the state of the art in 1985.
Again, simply a problem of technology and computing power. We can't do it today. Tomorrow, we'll be able to. The morality of it can be debated. I think we'll need new actors in recast roles, otherwise that career path dies like so many others are going to. Programs like Replicate can already create AI pictures almost indistinguishable from the real thing, so in the future it would be easy for studios to create entirely AI casts complete with AI derived voices. Just like a couple generations from now, guys like me (pilots) will be extinct, and air and ground vehicles will be beehive-networked and completely computer controlled. And I hate to say it, but when that technology matures, they'll be safer than they are today with humans behind the controls.
Progress. It's a real double-edged sword.
"TATV..."(ENT)
We're doomed.
We're doomed.
We're a loooong way off from Deep Fake or any machine learning tech being able to get an anywhere near convincing, emotive performance without a TON of bts manipulation, never mind basing a whole show around it. (Seriously, those thing can't even figure out how to draw hands yet!)
And depending on how the current strike plays out, what is and is not technically feasible 10 or 15 years down the line may be utterly irrelevant if it get expressly forbidden by the unions (and rightly so.)
I think people get a false perception of what really goes into producing something like this because all they see is the end result. "How did they do that?" they might ask. "Why, by the magic of deep fake of course!" replies the internet at large. "Wow, that's amazing! Just tighten up the graphics on level 3, and it'll be perfect and can do anything at the push of a button!"
Also let's not forget that an emotive performance is not something that can be programmed; I don't care how much processing power you throw at it. At best, all these things can do is mimic and iterate. An algorithm doesn't comprehend emotional context. At some point along the way, you need a human creative hand.
And depending on how the current strike plays out, what is and is not technically feasible 10 or 15 years down the line may be utterly irrelevant if it get expressly forbidden by the unions (and rightly so.)
I think people get a false perception of what really goes into producing something like this because all they see is the end result. "How did they do that?" they might ask. "Why, by the magic of deep fake of course!" replies the internet at large. "Wow, that's amazing! Just tighten up the graphics on level 3, and it'll be perfect and can do anything at the push of a button!"
Also let's not forget that an emotive performance is not something that can be programmed; I don't care how much processing power you throw at it. At best, all these things can do is mimic and iterate. An algorithm doesn't comprehend emotional context. At some point along the way, you need a human creative hand.
I don't agree -- if it's to be any good, rather than just generic pap, there will need to be human creativity involved, even if it's just a human tweaking the computer program to generate better stories and performances. Nothing is simply a problem of technology -- a tool is only as good as its wielder. The AI-generated "movies" we're getting now only look good to people who don't know what goes into making movies; actual filmmakers find them amateurish and superficial. It takes judgment to use any tool well.
Truly, the most compelling argument that there might actually be some human-like intelligence in these computer programs. Once computers become capable of drawing horses badly, we'll be in real trouble.
More seriously, though, the hand thing seems to have been largely fixed in the most recent releases in the past few months.
I'm actually pretty bullish on generative AI, possibly down to my own life experience. I decided I wanted to work with computer animation when I was a kid and one person could create a pretty good video game, a team of a dozen could do the visual effects for a TV show, and then entered the workforce in a world where you've got hundreds of artists all working infinitesimally small parts. A thousand years ago, I would've been a medieval carpenter who wants to build fancy chairs, but all anyone will pay for are cathedrals that'll take three hundred people three hundred years to finish. ChatGPT and Stable Diffusion are the first steps in how you get to a world where Tom Paris or the Doctor can create a perfect interactive narrative game single-handedly, or the woman in the Blade Runner sequel can make perfect photoreal animations with a device the size of a travel mug.
And both the creators and users of these programs are really fucking the dog on it, so I'm not really happy with how things are going, but I have faith that a balance will be found in the end and these things will become useful tools and not just vaguely-malignant toys that short-sighted people put the weight of the world on.
Though, to return to the subject, I'm still curious how much of Luke's fairly stilted performance in the new shows is because they're afraid of pushing the technology too far, or if they genuinely think that he should be a robotic prequel-Jedi who speaks exclusively in cover-letter, since people (by which I mean "people below the age of thirty on Reddit") love the prequel movies, where Jedi have a flat affect, and hate The Last Jedi, where this particular Jedi is sassy and emotive.
Come on, Luke! You liked Yoda, here's a baby, but he looks like Yoda. A Baby Yoda, if you will! He's cute as hell, you can crack a smile when you first see him. At least say your own name.
Maybe, but a lot of those people are going to be making things that they think are good but are still actually rubbish. When I started out as a writer, I thought I was writing good stories, but I needed five years' worth of rejection letters to learn just how inadequate they were and how much I needed to raise my game to be good enough for my work to be put before an audience in a professional publication.
There is something to be said for people having the freedom to create their own stuff for their own personal amusement and satisfaction, but if you want something good enough to have mass appeal for an audience, that's where talent, experience, and judgment come into play. What worries me is that if the technology makes it easy enough to churn out things that superficially look good, fewer people will bother to put in the work and discipline it takes to get genuinely good.
As I said; during testing for the BoBF episode they pushed it to destruction. So what we was in the final shots weren't the result of them being too afraid to push, it was the result them already having tried pushing and knowing exactly where it breaks (direct sunlight seemed to be a big one), thus going that far and no further. In other words, what we got was all it can do for the time being.
From what I've seen, they're mostly just for stealing other peoples work and passing it off as something new. Like microwaving a day old Big Mac and claiming to be a chef.
Speaking as someone that has a history in digital art (and stopped posting it publicly mostly because of wanton theft) I take a bit of a dim view of the idea that this is in anyway enabling for struggling artists. It's a shortcut for people without the talent, or patience to learn, who what to achieve their ambitions the easy way. It will put struggling artists out of work altogether. To borrow your analogy for a moment; what do you suppose the effect would be on the carpentry and stonemasonry business if all of a sudden, you don't need to hire 300 people to work for 300 years, but can instead 3000 randos are magicking 3000 cathedrals each out of thin air every day? Ever heard the term "demand collapse"?
Lets also not forget that these tools are by the admission of the people that made them utterly useless without a dataset to draw from. That dataset is images. Lots and lots of images done by other people who are not going to even be credited, let alone fairly compensated for they appropriated works. Because if they have to actually pay for what they're trying to profit from, suddenly the business model of the whole thing collapses.
You're right in that there will be an equilibrium at some point, and it'll happen shortly after someone in authority decides AI generated art by definition cannot be copyrighted, or even better, that it constitutes copyrighted theft just by existing. The investors will run a mile when they realise they can't just make a quick buck off the tech anymore. Indeed it's only a matter of time before someone like Disney realises someone is already feeding every Pixar movie ever into one of these things and can start churning out qualitatively indistinguishable knockoffs, that they can't collect licencing fees from, and they unleash the lawyers & lobbyists.
Last edited:
Again, I'm only talking about applying deep-fake 'looks' and 'voices' to motion-capture performances by human actors, acting out human-created scripts. Not a total reliance on AI for every level of the production.