I posted a longer comment above, but I am predicting that this is fairly close to as good as it gets in terms of AI video generation (using existing AI structures) without direct and substantial human intervention on a frame by frame basis The AI has no conception of what it is doing and will always diverge into chaos after a few seconds. There is no 'scene' being built by the AI, and I would wager that the most convincing and consistent movement in any AI animation is lifted directly from existing video sequences in which the density of pixels from one frame to the next diverge as little as possible and it will always tend towards chaos when that movement between frames becomes too extreme.
Very interesting to see how current text2video models interpret radically different concepts presented in retro-futuristim- albeit uncanny. The lighting and camera quality is way more spot-on than anything I've seen when modern shows or movies do flashbacks.
I think people are misinterpreting this in thinking your are trying to pass off AI generated video as legitimate retro futurism content. This is clearly just a showcase of the current technology and surreal horror of it.
I know that people have strong opinions on AI, and rightfully so, so I get some of the comments and criticism here but I appreciate you being able to see what I was going for!
@@picsnportraits Yeah I recently discovered that a good chunk of my artwork had been scrapped to train AI's you know, with ZERO permission. It's bad enough I have people in China, duplicating my posters and selling them and now robots are learning how to do my artwork so I won't even have a job to have artwork to get ripped off! Strong opinions is kinda an understatement.
Yeah, it’s awful this technology is built off of the exploitation of artists. I plan on talking about it more in depth in a future video but I don’t think creativity can be replicated. AI is only capable of copying, poorly, as seen in this video.
Some of these definitely have a nightmare quality to them. AI has a strange way animating, especially the dog that transforms into multiple things in the second clip
@@picsnportraits It's because the AI has no conception of the scene it is animating. It is purely interpolating frames based on the most likely next frame given the data it is trained on. It doesn't have any idea what is happening in the scene (it doesn't have any ideas at all) simply algorithmic allocation of pixels based on probabilities. I wouldn't be at all surprised to find that the most cohesive elements in any AI generated animations are simply consecutive frames from real videos (perhaps processed into a different style) that are being thrown together based on their high percentage of common pixels, since the movement between each frame is fairly minimal. You end up with short snippets of actual coherent movement copied directly from actual video clips which are quickly destroyed because the algorithm gets 'confused' when the step between frames in the source material becomes too large to tie preceding and proceeding frames together with enough certainty. Which is why the dog transforms into different images which are 'close enough' in pixel density to the original image of the dog but totally different in terms of their actual being. The algorithm was tripped up by its own data because the snippet of whatever footage the dog came from (or the combination of images which it was interpolating into one animated sequence of the dog) overlapped with another image in terms of pixel density. Since this 'AI' has no actual conception of what it is creating, other than matching pixel densities through complex sorting algorithms, it can't maintain coherence of scene or movement for more than a second or two that isn't essentially copied directly from existing sequences of images.
I know AI is controversial and this won’t become the norm. The intent here was to explore viewing the futures of the past through the future we are living in.
Love the surreal dream-like quality of these moving images, quite the strange sight.
I wonder if we’ll sentimentalize this early AI style similar to how we sentimentalize old art styles.
I was thinking that re: computer animation. So much early CGI is weird and janky but has a surreal, dreamlike quality to it like this does
I posted a longer comment above, but I am predicting that this is fairly close to as good as it gets in terms of AI video generation (using existing AI structures) without direct and substantial human intervention on a frame by frame basis
The AI has no conception of what it is doing and will always diverge into chaos after a few seconds. There is no 'scene' being built by the AI, and I would wager that the most convincing and consistent movement in any AI animation is lifted directly from existing video sequences in which the density of pixels from one frame to the next diverge as little as possible and it will always tend towards chaos when that movement between frames becomes too extreme.
Very interesting to see how current text2video models interpret radically different concepts presented in retro-futuristim- albeit uncanny. The lighting and camera quality is way more spot-on than anything I've seen when modern shows or movies do flashbacks.
i was both impressed and disturbed. the amount of 'uncanny valley' in that imagery was more than a little mind-bendy.
I think people are misinterpreting this in thinking your are trying to pass off AI generated video as legitimate retro futurism content. This is clearly just a showcase of the current technology and surreal horror of it.
I agree.
I know that people have strong opinions on AI, and rightfully so, so I get some of the comments and criticism here but I appreciate you being able to see what I was going for!
@@picsnportraits Yeah I recently discovered that a good chunk of my artwork had been scrapped to train AI's you know, with ZERO permission. It's bad enough I have people in China, duplicating my posters and selling them and now robots are learning how to do my artwork so I won't even have a job to have artwork to get ripped off! Strong opinions is kinda an understatement.
Yeah, it’s awful this technology is built off of the exploitation of artists. I plan on talking about it more in depth in a future video but I don’t think creativity can be replicated. AI is only capable of copying, poorly, as seen in this video.
i didnt see ONE comment about them thinking he was trying to pass these off as legitimate videos. where did you get that from?
0:05 this part is so incredible to look at first I thought it was from a movie then I noticed the dog turned into a girl . 😄
Hehehe, pure nightmare fuel, the one thing that visual AI is completely awesome at. It's not Will Smith eating spaghetti, but nicely done.
That's cool you animated the pictures, but that A.I generation was a bit terrifying!! Lol but really intriguing to watch
Some of these definitely have a nightmare quality to them. AI has a strange way animating, especially the dog that transforms into multiple things in the second clip
@@picsnportraits It's because the AI has no conception of the scene it is animating. It is purely interpolating frames based on the most likely next frame given the data it is trained on. It doesn't have any idea what is happening in the scene (it doesn't have any ideas at all) simply algorithmic allocation of pixels based on probabilities.
I wouldn't be at all surprised to find that the most cohesive elements in any AI generated animations are simply consecutive frames from real videos (perhaps processed into a different style) that are being thrown together based on their high percentage of common pixels, since the movement between each frame is fairly minimal. You end up with short snippets of actual coherent movement copied directly from actual video clips which are quickly destroyed because the algorithm gets 'confused' when the step between frames in the source material becomes too large to tie preceding and proceeding frames together with enough certainty. Which is why the dog transforms into different images which are 'close enough' in pixel density to the original image of the dog but totally different in terms of their actual being. The algorithm was tripped up by its own data because the snippet of whatever footage the dog came from (or the combination of images which it was interpolating into one animated sequence of the dog) overlapped with another image in terms of pixel density.
Since this 'AI' has no actual conception of what it is creating, other than matching pixel densities through complex sorting algorithms, it can't maintain coherence of scene or movement for more than a second or two that isn't essentially copied directly from existing sequences of images.
This was too freaky looking.
One of the coolest videos Ive seen on you tube. AI video is the future. Embrace it.
wow, your standards must be insanely low 😭
cool
Haunted!
Cool Video J
*Thanks!*
I've seen this movie
very cool video, nice and nostalgic, I like it
For the sake of humanity, and in memory of the artists who passionately crafted the media that you revisit here, please don't support generative AI.
Whoa, really don't like this. Hoping we don't see this a lot on the channel.
This was a fun experiment but will not be the norm!
This was actually really cool. I'd LOVE to see more.
Supremely unentertaining
Urgh not what I want to see on this channel....
I know AI is controversial and this won’t become the norm. The intent here was to explore viewing the futures of the past through the future we are living in.
@@picsnportraits You are 100% correct man, I am just being a negative nancy! Your channel is the best!
cool