I have worked on this video for a very long time, and I am happy to present you upscaled footage from around the world in 1896!Original footage from the Lumi...
Set to chill instrumental music.
Paris, Jerusalem, Istanbul, Geneva, Kyoto, London, Giza, New York, Germany, Madrid, Barcelona, Venice, Dublin, Moscow, Lyon, Marseille, etc.
Yeah, it looks pretty. But… I dunno. It wasn’t what was filmed at the time, and I sort of like that. What we are seeing is gaps filled in and made up by computers, a fiction to make us happy, no more truth than in the original.
But yeah, still cool even if not strictly history.
I mean, 60fps from 18fps. That’s what 42 frames per second “interpolated”. That means two thirds of this, as content, are inferred rather than being primary evidence.
Using the terms “low quality” and “high quality” are odd too. Subjective modernisms. Sure, that’s fine when denoting something purely as entertainment, but it doesn’t hold up as fact.
I don’t know why this irks me so much, I guess it shouldn’t.
I think you’re right to be wary of it. It’s like AI colorization of photos. They tend to be much more desaturated than ones done by professionals. Great colorists take the time to research materials and fashions and make informed guesses about how things should look. When something automated by AI it can be much more inaccurate since no research was done.
I would imagine the same pitfalls could happen with a video like this. It’s cool to see the results, but I agree it shouldn’t be presented as anything more than an artistic re-interpretation.
Yeah I once saw a “colorization” done by AI from a photo from the early 1900s or something, it was a photo from a house and some people in front of it, Russia or Ukraine somewhere.
The thing is that the photo that was used was in fact a color photo that they turned black and white and then gave it to the AI. The AI used very muted colors, a little beige a little dark green - colors you would expect from a photo of that time and region. In the original you could see, that the colors where actually very bright and plentiful, the house was a bright yellow not beige, the green fence was not dark green but bright green, the accents on the house were not brown but red.
The people in the AI colorized picture had black, white, dark blue clothes, a man had a brown hat, in the original they wore blue dresses, yellow blouses, the hat was red and so on.
Of course you can extrapolate some things but the original photo was so colorful, these people clearly painted the house bright on purpose and they wore very bright clothes but the re-colorization was so much less colorful and it gave a wrong impression of the reality what was photographed. Of course the old film wasn’t perfect and the colors maybe were also a little off - but it showed how many different colors were present in the original.
So I think in this case the motion interpolation can legitimately increase the verisimilitude of the footage. The pace and fluidity of the movement being more natural is not a bad thing for the showing off the times, though it is important that it be noted it’s a particularly poor reflection on the history of cinema and its cultural impact. The two thirds that are missing have the 1/18 of a second on either side there, so I think there’s a particularly low risk in introducing misleading information.
My bigger concerns are actually the upscaling (a bit) and the colorization (more). The former, I guess if you’re just sort of presenting this to create the impression of these people’s faces and to enhance the immersion in the era for a modern audience, it’s not that bad, but you’d want to be very clear what you were doing, and you certainly wouldn’t want to say something like, “See what your great-grandfather’s face really looked like?” For the colorization, I’d want to know what were the sources, techniques, and tools used. Those would befit from genuine historical research and could be actively misleading about what we’re seeing, providing a false certainty in a way that motion interpolation mostly doesn’t, and upscaling sort of doesn’t.
Yeah. I think that’s fair. Although I’m interested to know why they chose 60fps rather than 24. It seems like a flex as opposed to a genuine desire to show fluid movement.
I suspect that, in a very pure way, the study of the colourisation could be an interesting academic pursuit that would reveal more about what we are looking at. Though that would require a ton of work and would still require a fair amount of presumption to be “complete”.
But there’s the rub. “Modern audiences”. Rather than pander to an expectation that things have to look a particular way now surely we should encourage people to see how it was recorded then?
The very fact there is film documentation of a scene in 1896 is interesting in its own right, and for want of a better phrase, it is what it is. This is what footage from over a hundred years ago looks like. I guess I’m not that comfortable with a revisionist history of media.
Not sure what that’s got to do with this… I mean you are right, there’s a lot of editing going on with smartphone images, but procedural and untentional. The current scandal over the princess of Wale’s clearly photoshopped family portrait is indicative.
But just because one technology messes with things doesn’t provide an excuse for that to happen across the board.
And I’m not being romantic, I’m talking about the veracity of primary historic documentation vs the need for someone to see something in colour at 60fps.
Most pictures are postprocessed automatically nowadays. Beauty filters, pixel binding, HDR…
Multiple layers and filters which distort ‘reality’.
And I think you are being romantic by trying to preserve a form of originality, while technology itself has never done anything else than alter our perception of the world.
As an example: There’s a good reason why many kids might have thought that grandpa’s reality looked black and white back then haha
How much something resembles reality in this case is subjective.
Some might even argue the AI footage looks closer to the truth than the original file.
Yeah, it looks pretty. But… I dunno. It wasn’t what was filmed at the time, and I sort of like that. What we are seeing is gaps filled in and made up by computers, a fiction to make us happy, no more truth than in the original.
But yeah, still cool even if not strictly history.
I mean, 60fps from 18fps. That’s what 42 frames per second “interpolated”. That means two thirds of this, as content, are inferred rather than being primary evidence.
Using the terms “low quality” and “high quality” are odd too. Subjective modernisms. Sure, that’s fine when denoting something purely as entertainment, but it doesn’t hold up as fact.
I don’t know why this irks me so much, I guess it shouldn’t.
I think you’re right to be wary of it. It’s like AI colorization of photos. They tend to be much more desaturated than ones done by professionals. Great colorists take the time to research materials and fashions and make informed guesses about how things should look. When something automated by AI it can be much more inaccurate since no research was done.
I would imagine the same pitfalls could happen with a video like this. It’s cool to see the results, but I agree it shouldn’t be presented as anything more than an artistic re-interpretation.
Yeah I once saw a “colorization” done by AI from a photo from the early 1900s or something, it was a photo from a house and some people in front of it, Russia or Ukraine somewhere.
The thing is that the photo that was used was in fact a color photo that they turned black and white and then gave it to the AI. The AI used very muted colors, a little beige a little dark green - colors you would expect from a photo of that time and region. In the original you could see, that the colors where actually very bright and plentiful, the house was a bright yellow not beige, the green fence was not dark green but bright green, the accents on the house were not brown but red. The people in the AI colorized picture had black, white, dark blue clothes, a man had a brown hat, in the original they wore blue dresses, yellow blouses, the hat was red and so on.
Of course you can extrapolate some things but the original photo was so colorful, these people clearly painted the house bright on purpose and they wore very bright clothes but the re-colorization was so much less colorful and it gave a wrong impression of the reality what was photographed. Of course the old film wasn’t perfect and the colors maybe were also a little off - but it showed how many different colors were present in the original.
So I think in this case the motion interpolation can legitimately increase the verisimilitude of the footage. The pace and fluidity of the movement being more natural is not a bad thing for the showing off the times, though it is important that it be noted it’s a particularly poor reflection on the history of cinema and its cultural impact. The two thirds that are missing have the 1/18 of a second on either side there, so I think there’s a particularly low risk in introducing misleading information.
My bigger concerns are actually the upscaling (a bit) and the colorization (more). The former, I guess if you’re just sort of presenting this to create the impression of these people’s faces and to enhance the immersion in the era for a modern audience, it’s not that bad, but you’d want to be very clear what you were doing, and you certainly wouldn’t want to say something like, “See what your great-grandfather’s face really looked like?” For the colorization, I’d want to know what were the sources, techniques, and tools used. Those would befit from genuine historical research and could be actively misleading about what we’re seeing, providing a false certainty in a way that motion interpolation mostly doesn’t, and upscaling sort of doesn’t.
Yeah. I think that’s fair. Although I’m interested to know why they chose 60fps rather than 24. It seems like a flex as opposed to a genuine desire to show fluid movement.
I suspect that, in a very pure way, the study of the colourisation could be an interesting academic pursuit that would reveal more about what we are looking at. Though that would require a ton of work and would still require a fair amount of presumption to be “complete”.
But there’s the rub. “Modern audiences”. Rather than pander to an expectation that things have to look a particular way now surely we should encourage people to see how it was recorded then?
The very fact there is film documentation of a scene in 1896 is interesting in its own right, and for want of a better phrase, it is what it is. This is what footage from over a hundred years ago looks like. I guess I’m not that comfortable with a revisionist history of media.
Romantic, but lemme introduce you to smartphone photography!
Not sure what that’s got to do with this… I mean you are right, there’s a lot of editing going on with smartphone images, but procedural and untentional. The current scandal over the princess of Wale’s clearly photoshopped family portrait is indicative.
But just because one technology messes with things doesn’t provide an excuse for that to happen across the board.
And I’m not being romantic, I’m talking about the veracity of primary historic documentation vs the need for someone to see something in colour at 60fps.
Most pictures are postprocessed automatically nowadays. Beauty filters, pixel binding, HDR…
Multiple layers and filters which distort ‘reality’.
And I think you are being romantic by trying to preserve a form of originality, while technology itself has never done anything else than alter our perception of the world.
As an example: There’s a good reason why many kids might have thought that grandpa’s reality looked black and white back then haha
How much something resembles reality in this case is subjective.
Some might even argue the AI footage looks closer to the truth than the original file.
Bah humbug
The original scenes didn’t happen in black and white. I’d argue this is even more accurate to the original (real life).
Yeah, that’s not how primary evidence works.