@@AMA14700 Isn't is live action already? I mean like, different definitions of "live action", but all these movies people have been decrying were all called live action even with a cgi as opposed to practical effect backbone
At 7:55 it sounds like you say 'ass to mouth' and that mixed with the moustache guy onscreen is just hilarious. Unintended jokes aside, though, this is a great summary of what's around at the moment and I often find myself sharing your vids when I'm trying to explain some of this stuff quickly to people I'm talking about it with. Thanks
Haven't watched the whole video yet but you joking about the black box and Chinese engineers made me think of the Chinese room problem but it's just a random chance that there's just some random Chinese Ph.D student inside arranging the characters.
Have to admit a video about AI generated a video that has a video to video ad in its timeline is pretty clever. Also where the category of 1950s Panavision video seems like that should be included simply based on the sheer amount.
I liked your videos but you didn't named the tools that achievs that kind of results. For example the spaghetti dancing ai video was made in stable diffusion animatediff. So I would like you to make a video on tools also.
Great way to categorize the models and their use cases. I'm yet to find a tool/model for image manipulation using gen AI. You give it a source image and you can modify it or add elements to it using a prompt (not just as a reference for an image generation).
@@MangaGamified that would work if I need to add an element to an unchanged image. But, that wouldn't work if I want to make seamless changes to multiple elements of the image, its style, or composition using a prompt.
@@T33KS Sounds like it will take the same amount of time and effort if you prop up assets and scenes in blender, render them over a greenscreen/transparent background with a shadow catcher. Place them over your desired image. I'm not hating AI, I use it for thumbnailing or storyboarding, or as place holders, but as a customer, client or viewer it's just feels lazy or betrayal. Even CGI feelz lazy even if you model, subdivide, decimate, project high details on low-poly, uv-unwrap, texture, bake, etc them. or the classic, paint everthing by hand if not everything is a 3D asset in blender. Lastly, don't under estimate humans perceptions, sometimes we're just too good at perception if it's fake(if we're not too gullible nor drunk at that time. )
@@MangaGamified I agree that with today's tools and models, AI generated is still perceivable and frowned upon by most clients (and many artists). I say "most clients" because at the other end of the spectrum are those who want their content to be explicitly AI generated. They're not trying to fool the audience. On the contrary, they want to be percieved as avantgarde and up-to-date. AI's impact on media is huge, and it will get even bigger. Love it or hate it, you can't stop the train now.
Real video AI is actually the easiest, all you need to do is feed it any, non-edited, real world capture and train it ¯\_(ツ)_/¯ There is no bad quality video data
Every type of AI generated video ever, explained provides a comprehensive overview of various AI-generated video techniques, including deepfakes and automated video editing. It clarifies how these technologies work and their implications for media and content creation.
As far as I can tell there isn't and won't be enough data to actually be a threat to film and animation, especially animation. They'd basically need to produce thousands or hundreds of thousands of training data of animation by which point you might as well just... make the films.
@@olivetree9920 Unfortunately I don't remember the exact video, but it was a scientist who read a bunch of metanalysis on the progress of AI. Most of the current data appears to show somewhat of an slowdown. At the same time there's apparently 2 camps at the moment: Those who think like I do on the subject and those who think there's no limit and we are just getting warmed up and what we've seen so far is nothing compared to what's coming.
no we are far from hitting the data wall for video, that's only the case with text (and perhaps images) and for text there is synthetic data (with verifiers to prevent model collapse). Video data is abundant, the main limiter is algorithm and compute.
its not like all of the data has to be handdrawn when it can be mostly cgi with different filters. combined with models that can effectively control skeletons, it might be possible to mass produce low effort but instruction-adherent animations for pretraining
check out DomoAI and start your vid2vid experiments: r.domoai.app/bycloud
If you enjoy these types of compilations, let me know!
AI generated minecraft movie when?
In... Two seconds.
At least it won't be live action 😂😂😂😂
@@AMA14700 Isn't is live action already? I mean like, different definitions of "live action", but all these movies people have been decrying were all called live action even with a cgi as opposed to practical effect backbone
I mean. It already is.
A slop.
@@FaultyTwo honestly yeah
You should do a video on what is AI and when it is just linear regression.
At 7:55 it sounds like you say 'ass to mouth' and that mixed with the moustache guy onscreen is just hilarious.
Unintended jokes aside, though, this is a great summary of what's around at the moment and I often find myself sharing your vids when I'm trying to explain some of this stuff quickly to people I'm talking about it with. Thanks
Haven't watched the whole video yet but you joking about the black box and Chinese engineers made me think of the Chinese room problem but it's just a random chance that there's just some random Chinese Ph.D student inside arranging the characters.
@@mcmann7149 pretty much
AI anime when?
Next week
Soon™ but it only generates NTR
@@MangaGamified let it be something
Great video today! Obviously, right up my alley on this one!
Have to admit a video about AI generated a video that has a video to video ad in its timeline is pretty clever. Also where the category of 1950s Panavision video seems like that should be included simply based on the sheer amount.
@@southcoastinventors6583 you lost me at "Have"
@@BadChess56 One response is better than none
I liked your videos but you didn't named the tools that achievs that kind of results. For example the spaghetti dancing ai video was made in stable diffusion animatediff. So I would like you to make a video on tools also.
Thank you for sharing really informative. I enjoyed it a lot
Thanks
Great way to categorize the models and their use cases.
I'm yet to find a tool/model for image manipulation using gen AI.
You give it a source image and you can modify it or add elements to it using a prompt (not just as a reference for an image generation).
I think you can just prompt for an object, crop it, have a transparent background, and export it as PNG, then paste over your image
@@MangaGamified that would work if I need to add an element to an unchanged image. But, that wouldn't work if I want to make seamless changes to multiple elements of the image, its style, or composition using a prompt.
@@T33KS Sounds like it will take the same amount of time and effort if you prop up assets and scenes in blender, render them over a greenscreen/transparent background with a shadow catcher. Place them over your desired image.
I'm not hating AI, I use it for thumbnailing or storyboarding, or as place holders, but as a customer, client or viewer it's just feels lazy or betrayal. Even CGI feelz lazy even if you model, subdivide, decimate, project high details on low-poly, uv-unwrap, texture, bake, etc them.
or the classic, paint everthing by hand if not everything is a 3D asset in blender.
Lastly, don't under estimate humans perceptions, sometimes we're just too good at perception if it's fake(if we're not too gullible nor drunk at that time. )
@@MangaGamified I agree that with today's tools and models, AI generated is still perceivable and frowned upon by most clients (and many artists).
I say "most clients" because at the other end of the spectrum are those who want their content to be explicitly AI generated. They're not trying to fool the audience. On the contrary, they want to be percieved as avantgarde and up-to-date.
AI's impact on media is huge, and it will get even bigger. Love it or hate it, you can't stop the train now.
4:56 Yep,, me in my hallway **
Step by step closers to AI-FDVR Lorapilled Animemaxxing.
Cant wait for the wave of guys with female avatars
Guess you never played a MMORPG
Monster Hunter in a nutshell.
Can't*
Literally working on this right now, just what I needed!
Real video AI is actually the easiest, all you need to do is feed it any, non-edited, real world capture and train it ¯\_(ツ)_/¯ There is no bad quality video data
You need to label the video that you feed in. Otherwise you would not get what you asked for when you promt
@@enkvadrat_ Duh, but that also occurs with other datasets. In this case you can save a lots of filtering out bad datasets.
What if the video is static?
@@plaiday Not an issue, the model has to learn that not everything in a video moves.
Every type of AI generated video ever, explained provides a comprehensive overview of various AI-generated video techniques, including deepfakes and automated video editing. It clarifies how these technologies work and their implications for media and content creation.
Calling it Nick Avocado is using body swap 😅
we need explanation whether nikocado is real or not
Ai generated skibidi toilet when?
Soo cool, so damn dangerous
this stuff has the same use-case potential as ico projects did in 2017. at some point we’ll accept that it’s best for internet goofing and scams.
🔥 vid fr fr❌️🧢
We ❤🔥 N✝R fr fr
As far as I can tell there isn't and won't be enough data to actually be a threat to film and animation, especially animation. They'd basically need to produce thousands or hundreds of thousands of training data of animation by which point you might as well just... make the films.
Where did you get that info?
@@olivetree9920 Unfortunately I don't remember the exact video, but it was a scientist who read a bunch of metanalysis on the progress of AI. Most of the current data appears to show somewhat of an slowdown. At the same time there's apparently 2 camps at the moment: Those who think like I do on the subject and those who think there's no limit and we are just getting warmed up and what we've seen so far is nothing compared to what's coming.
no we are far from hitting the data wall for video, that's only the case with text (and perhaps images) and for text there is synthetic data (with verifiers to prevent model collapse). Video data is abundant, the main limiter is algorithm and compute.
its not like all of the data has to be handdrawn when it can be mostly cgi with different filters. combined with models that can effectively control skeletons, it might be possible to mass produce low effort but instruction-adherent animations for pretraining
If you don't see any improvement in the next 10 years you are right but so far ...
Boring