Ok, maybe I am too comfortably nested in my confirmation bubble, but I can’t imagine the intersection of people being (still) interested in Orson Welles and those not offended by such a use of AI is very large.
I mean, it’s not like I’ve been waiting for this my whole life, or anything, but I’ve watched around lost media in a lot of cases, from lost Doctor Who episodes to different restorations and partial cuts of Nosferatu. Reconstructions of lost media often do the best they can with what they have, it’s never perfect. Both fan and professional animated reproductions of lost Who episodes where only audio and pictures persist are often janky and poorly directed (then again, so is the original material). We don’t know what the original cut of Nosferatu actually was, despite the hard work of very legit scholars and historians.
Ambersons has a lot more metadata around it about the lost material than many other equivalents. It’s probably easier to rebuild through any means than other things. What tools you use to get there is not paticularly concerning to me, and it will always be a curio hybrid product, no matter what you use to put it together. In my mind it sits somewhere between using test footage to sort of cobble together the Donner cut of Superman II and… well, I’d love to see the original cut of The Private Life of Sherlock Holmes, but that seems like it’d involve more guesswork than this.
Ultimately, like everything else, it depends a lot on how knowledgable, skilled and respectful the people working on the restoration are. That has nothing to do with how they generate any materials that aren’t directly preserved footage.
You saying about this film having a lot of documentation around the lost 43 minutes made me look into it. I did not know the story behind it, i.e. it being already cut by Welles, then the 43 minutes being cut out by studios, plus a lot of research and reconstruction already being made around it. Adding to that the fact this is not (thus far) a commercial endeavour, it does paint it in a different light. Finally, from what I can gather, it seems the “AI” being used here is more deepfake stuff on live scenes and less full image generation (which is the image that the text conjured for me, this is the problem with catchall marketing terms…)
All that to say, while I personally am not into these kinds of efforts (AI or not, but I appreciate the subjectivity of that sentiment), and have my reservations about using these techs to reanimate long dead artists who don’t have a say in the matter, your comment did show that the process, in this particular instance, seems to be very different from what I had initially imagined, so thank you.
Sorry about the downvotes and potentially angry responses you are/will be getting, I did not mean to lay down a trap for you.
Man, I do appreciate the sentiment, but I’m an AI-agnostic guy poking at the Fuck AI community to argue about nuance. You are 100% not to blame for any hate coming my way.
Which is relatively little, for the record, because this is still Fedi and most people here are somewhere between civilized and quite nice.
I genuinely don’t have a good idea of what they’ll try to do and how. There’s certainly been this thing where any machine learning tool at all gets hate now, regardless of whether it’s being used to resolve pictures of black holes, restore old movies or… whatever people do with generative AI in their day to day. People are mostly out to have a good ole dogpile because it’s a feel-good angle of anticorporate activism you can do at home.
And I get it, I’m just too much of a contrarian to throw babies out with the bathwater when you can have a spirited discussion that leads nowhere instead. Glad it had any positive impact on anybody at all, honestly.
I am not, in fact, using tools of any kind to recreate the lost director’s cut of Orson Wells’ 1942 classic The Magnificent Ambersons, so I have nothing to stop.
I’m going to guess the answer you’d get from the Gen AI company that is actually doing it as a promotional/academic action would be “No”.
This seems reasonable.
As for myself, I don’t use any online LLMs or generative AI, but I do sometimes mess around with local open source models out of curiosity and for some applications where it makes sense.
My answer is also “No”. Maybe just a little bit harder of a no now than before reading this.
You’re doing a bang-up job. Truly winning hearts and minds out here.
Ok, maybe I am too comfortably nested in my confirmation bubble, but I can’t imagine the intersection of people being (still) interested in Orson Welles and those not offended by such a use of AI is very large.
Hi, it me.
I mean, it’s not like I’ve been waiting for this my whole life, or anything, but I’ve watched around lost media in a lot of cases, from lost Doctor Who episodes to different restorations and partial cuts of Nosferatu. Reconstructions of lost media often do the best they can with what they have, it’s never perfect. Both fan and professional animated reproductions of lost Who episodes where only audio and pictures persist are often janky and poorly directed (then again, so is the original material). We don’t know what the original cut of Nosferatu actually was, despite the hard work of very legit scholars and historians.
Ambersons has a lot more metadata around it about the lost material than many other equivalents. It’s probably easier to rebuild through any means than other things. What tools you use to get there is not paticularly concerning to me, and it will always be a curio hybrid product, no matter what you use to put it together. In my mind it sits somewhere between using test footage to sort of cobble together the Donner cut of Superman II and… well, I’d love to see the original cut of The Private Life of Sherlock Holmes, but that seems like it’d involve more guesswork than this.
Ultimately, like everything else, it depends a lot on how knowledgable, skilled and respectful the people working on the restoration are. That has nothing to do with how they generate any materials that aren’t directly preserved footage.
Thanks for this answer, really.
You saying about this film having a lot of documentation around the lost 43 minutes made me look into it. I did not know the story behind it, i.e. it being already cut by Welles, then the 43 minutes being cut out by studios, plus a lot of research and reconstruction already being made around it. Adding to that the fact this is not (thus far) a commercial endeavour, it does paint it in a different light. Finally, from what I can gather, it seems the “AI” being used here is more deepfake stuff on live scenes and less full image generation (which is the image that the text conjured for me, this is the problem with catchall marketing terms…)
All that to say, while I personally am not into these kinds of efforts (AI or not, but I appreciate the subjectivity of that sentiment), and have my reservations about using these techs to reanimate long dead artists who don’t have a say in the matter, your comment did show that the process, in this particular instance, seems to be very different from what I had initially imagined, so thank you.
Sorry about the downvotes and potentially angry responses you are/will be getting, I did not mean to lay down a trap for you.
Man, I do appreciate the sentiment, but I’m an AI-agnostic guy poking at the Fuck AI community to argue about nuance. You are 100% not to blame for any hate coming my way.
Which is relatively little, for the record, because this is still Fedi and most people here are somewhere between civilized and quite nice.
I genuinely don’t have a good idea of what they’ll try to do and how. There’s certainly been this thing where any machine learning tool at all gets hate now, regardless of whether it’s being used to resolve pictures of black holes, restore old movies or… whatever people do with generative AI in their day to day. People are mostly out to have a good ole dogpile because it’s a feel-good angle of anticorporate activism you can do at home.
And I get it, I’m just too much of a contrarian to throw babies out with the bathwater when you can have a spirited discussion that leads nowhere instead. Glad it had any positive impact on anybody at all, honestly.
Or, you could not use the plagiarism and theft machine to hallucinate missing content.
That is quickly becoming the least useful thing to say on the Internet.
Still infinitely more useful than “I asked ChatGPT” or any other lazy LLM.
Sure, man. Look, if you’re just going to repeat slogans out of context I promise you I’ve heard them. You can save yourself the typing.
Deal with it or stop using the plagiarism and theft machine.
I am not, in fact, using tools of any kind to recreate the lost director’s cut of Orson Wells’ 1942 classic The Magnificent Ambersons, so I have nothing to stop.
I’m going to guess the answer you’d get from the Gen AI company that is actually doing it as a promotional/academic action would be “No”.
This seems reasonable.
As for myself, I don’t use any online LLMs or generative AI, but I do sometimes mess around with local open source models out of curiosity and for some applications where it makes sense.
My answer is also “No”. Maybe just a little bit harder of a no now than before reading this.
You’re doing a bang-up job. Truly winning hearts and minds out here.