I think it's quite disingenuous to dismiss the lawsuits on the basis that the models aren't violating copyright, and that rather it's the scraping of data sets that is. These models would be useless without the tens of thousands of images whose copyright was ignored to make the latent images. This is especially true when the creators of the models couldn't have made a dataset to train it themselves for lack of talent. Also AI art enthusiasts can be equally as militant as it's detractors
Need a blooper real at the end of the anime episode! Seeing the character breaks after cutting sounds wild. Also really love listening to you guys talk about the future of ai, best podcast!
Niko, love ya man but you should be putting your actions where your mouth is then for utilizing Stable Diffusion. Find an ethically trained model, then use that for your work. Boom, done. Ethical backlash is done and you're covered.
Great episode! I'm really excited for this video. Like Dean, Corridor's anime videos have always been my favorites since Anime Baseball came out. I regularly quote Anime Fidget Spinners in my daily life. I'm what you might refer to as a "Stale Trueton", been watching since 2011. So any new narrative uploaded to the Corridor channel is a treat for me.
If you don't have a good story or a message to communicate to the audience how do you think you can engage people to watch your films? Even if the final product is well done technically. From my point of view there has to be a balance between content and form.
A person I know was working on video-motion capture in like 2020 as part of their PhD research, originally looking at capturing the skeleton/motion of ice skaters. You could use the research to map the skeleton/motion and transplant that motion onto the models?
working at the speed of thought... that's the term I've been looking for. Atm 3d animation is so so so so far from that, pains me everyday! can't wait for this A.I stuff to get really good
just a point to their claim that this tech wont steal jobs. they say that they would not have hired anyone to make their stuff which is fair, but my boss did hire a bunch of people, then fired them all and hired people in lower cost of living places to do the same work slightly cheaper/shittier cause it makes him more money. people like him are salivating over this tech at the thought of not having to pay artists anymore
11:30 but that is what is going to be used to set precedent. So it doesn't matter that just "a" model is being hit for copyright. It means that everyone after will be taken out easily
It's hilarious looking back in retrospect because I was one who thought differently about the ai art stuff until I saw your video and you guys giving a better perspective than the bs every news outlet or Instagram copy-paste posts
You guys should watch Naruto and Boruto fights on youtube. The fighting cohegraphy is so cool, it really inspiring! It is beyond what anyone could film, it's so much more creative !!
I hope you guys contacted James Lee, if you aim to making live action footage into anime. I feel like he would have some valuable knowledge and tips on how to retime real-life motion into stylized motion. Because I'm afraid that if you just plop live action footage into AI - it'll look like those rotoscopped movies - cool looking still, but not exactly "like anime". Anime has a ton of stylized movement.
I use image to image all the time, sometimes I just take a picture of random junk and squint my eyes and think to myself, what does that look like, then that informs my prompt, I'm also doing a similar thing with blender AI render, using simple Geo and assets to figure out the comp and then drop in a prompt and let the magic begin
Looking forward to this. I've got a green screen and software, I'd love to have a crack at this tech. Although my son would just be like Lets make a Siren Head anime! Yeah no.
Legal Eagle's coverage of this topic is interesting, if you guys haven't watched it. The latent space transformation step of training may be argued by the plaintiff(s) as a derivative work by definition. Should be interesting to watch the legal space around it develop. Plus, even if we wind up with a court system that sides with the artists, at least we'll get to watch someone train a model using massive databases of court stenography, then train a GAN specifically to find the best arguments to win copyright cases. The job of the model trainers is secure.
My idea was to make a datamoshed video with no i-frames, to warp the noise along with the video motion, then image to image every frame with that as the noise source My gpu can't run stablediffusion or I'd try it myself lol
Every time you talk about ai art you sound so privileged. You understand the impact on paper, but you still don't truly get how it feels for software to show up in a matter of months and try to erase something you held dear and worked really hard to achieve. You've been on the receiving and of these tech benefits the entire time and it shows. Every time a new piece of tech comes along, it makes your passion, your art, easier. Faster. Better. But that's how it starts. When the wave passes you, and it's your lunch being eaten, you'll understand but it'll be too late. Technological progress is dependent on removing inefficiency. What do you think happens when the least efficient part of the system is you?
I think Corridor Cast doesn't seems to be understand implication of the generative AI. For them it's tool to make more cheap contents. But the real issue is the future of the art as a whole. It cheapens the art into more generic stuff gives even more bad smelly stigma around it, basically driving away, stunting new generation of artists. It means more tech people creating generic shits, with no vision, creative direction and basically levels down for generations. There is no winning party of this AI generative arts unless it's cheap content makers, disillusioned ai art bros. It's stealing art and top of that even money they making, lively hood.
Fascinating I'm watching brilliant people through Internet, on my computer at home, with them talking about awesome innovations (in a recorded podcast - not just written down ideas) in movie and television video graphics with the aid of computers and now with state-of-the-art AI image generation..... what a time to be alive truly. So happy and thankful to be a fan of Corridor Digital.
Really wish you guys did more of this pod, it's so dope.
I think it's quite disingenuous to dismiss the lawsuits on the basis that the models aren't violating copyright, and that rather it's the scraping of data sets that is. These models would be useless without the tens of thousands of images whose copyright was ignored to make the latent images. This is especially true when the creators of the models couldn't have made a dataset to train it themselves for lack of talent.
Also AI art enthusiasts can be equally as militant as it's detractors
I listen on Apple Podcasts, but commenting here for that YT algorithm goodness. Keep it up my dudes!
Need a blooper real at the end of the anime episode! Seeing the character breaks after cutting sounds wild. Also really love listening to you guys talk about the future of ai, best podcast!
Niko, love ya man but you should be putting your actions where your mouth is then for utilizing Stable Diffusion. Find an ethically trained model, then use that for your work. Boom, done. Ethical backlash is done and you're covered.
Can I introduce you to RPS Anime 2? 😉
Great episode! I'm really excited for this video. Like Dean, Corridor's anime videos have always been my favorites since Anime Baseball came out. I regularly quote Anime Fidget Spinners in my daily life.
I'm what you might refer to as a "Stale Trueton", been watching since 2011. So any new narrative uploaded to the Corridor channel is a treat for me.
*Jake is not there*
Sam and Niko: *ANIMEEEEEE*
If you don't have a good story or a message to communicate to the audience how do you think you can engage people to watch your films? Even if the final product is well done technically. From my point of view there has to be a balance between content and form.
This reminds me of that movie A Scanner Darkly.
You could totally do that to any movie now.
A person I know was working on video-motion capture in like 2020 as part of their PhD research, originally looking at capturing the skeleton/motion of ice skaters. You could use the research to map the skeleton/motion and transplant that motion onto the models?
working at the speed of thought... that's the term I've been looking for. Atm 3d animation is so so so so far from that, pains me everyday! can't wait for this A.I stuff to get really good
Looking forward to Wrens next high energy 'Scale-of' video with the back of Sam's super chill head in the background lol
just a point to their claim that this tech wont steal jobs. they say that they would not have hired anyone to make their stuff which is fair, but my boss did hire a bunch of people, then fired them all and hired people in lower cost of living places to do the same work slightly cheaper/shittier cause it makes him more money. people like him are salivating over this tech at the thought of not having to pay artists anymore
It's pretty stupid to think this tech won't replace jobs.
11:30 but that is what is going to be used to set precedent. So it doesn't matter that just "a" model is being hit for copyright. It means that everyone after will be taken out easily
It's hilarious looking back in retrospect because I was one who thought differently about the ai art stuff until I saw your video and you guys giving a better perspective than the bs every news outlet or Instagram copy-paste posts
Vtubers might have been doing what you're hinting at?
You guys should watch Naruto and Boruto fights on youtube. The fighting cohegraphy is so cool, it really inspiring! It is beyond what anyone could film, it's so much more creative !!
We're missing the "Jaime" of this podcast, someone who could bring it up and show us what Sam is talking about it.
I have seen animated characters breaking character after a director saying cut in an animated series where they animated bloopers
At the end there Sam is bursting at the seams trying to hold in all the things he wants to call the audience. Let it out!
CorridorCast: *proceeds as norman*
Wren in the background: A N T S
Its just the same argument in a way of traditional brush and canvas artists vs photoshoppers from years past.
It is very selfish to dismiss the concerns of people who had their work misappropriated to create software.
my twitter feed is really curated too, mostly art posts and minimal text
most of the ads are those goofy mobile ads too, so thats a bonus
Sucker Punch is as near as dammit live action anime style!!
Nico says at 13:00 though that the only thing that matters is the concept haha
it's interesting not more than that...
What on Earth is the music used in the ridgewallet ad??!!
You guys really gotta watch Redline.
Toy Story did animated bloopers at the end of the movie. This is something I’m surprised we don’t see more of.
Toy Story, A Bug's Life, and Cars had blooper reels, and I think there are others, but I don't remember.
Hold up, did I just hear Dubstep Guns 2?
Couldn't you just make a new model from the output of the first?
Yes you can for the model they used they also did some of that because there where issues with the beard of niko
I hope you guys contacted James Lee, if you aim to making live action footage into anime. I feel like he would have some valuable knowledge and tips on how to retime real-life motion into stylized motion. Because I'm afraid that if you just plop live action footage into AI - it'll look like those rotoscopped movies - cool looking still, but not exactly "like anime". Anime has a ton of stylized movement.
I use image to image all the time, sometimes I just take a picture of random junk and squint my eyes and think to myself, what does that look like, then that informs my prompt, I'm also doing a similar thing with blender AI render, using simple Geo and assets to figure out the comp and then drop in a prompt and let the magic begin
If the ridge wallet has rfid blocking technology how did Jordan access the studio with her access card inside the wallet 🤔
Looking forward to this. I've got a green screen and software, I'd love to have a crack at this tech. Although my son would just be like Lets make a Siren Head anime! Yeah no.
Man I'm so hyped for this video!!!
7:44
When you release this new anime project can you also release a blooper reel. Animated character can now have real bloopers.
Legal Eagle's coverage of this topic is interesting, if you guys haven't watched it. The latent space transformation step of training may be argued by the plaintiff(s) as a derivative work by definition. Should be interesting to watch the legal space around it develop. Plus, even if we wind up with a court system that sides with the artists, at least we'll get to watch someone train a model using massive databases of court stenography, then train a GAN specifically to find the best arguments to win copyright cases. The job of the model trainers is secure.
Woops didn't realize you guys covered this last week lol
Stoked the podcasts are back again!
My idea was to make a datamoshed video with no i-frames, to warp the noise along with the video motion, then image to image every frame with that as the noise source
My gpu can't run stablediffusion or I'd try it myself lol
Super excited for this episode!
Chainsaw man in real life?
Sadly that’s impossible because of the sad fact that pochita doesn’t exist irl and chainsawman can’t exist whiteout pochita
Every time you talk about ai art you sound so privileged. You understand the impact on paper, but you still don't truly get how it feels for software to show up in a matter of months and try to erase something you held dear and worked really hard to achieve. You've been on the receiving and of these tech benefits the entire time and it shows. Every time a new piece of tech comes along, it makes your passion, your art, easier. Faster. Better. But that's how it starts. When the wave passes you, and it's your lunch being eaten, you'll understand but it'll be too late.
Technological progress is dependent on removing inefficiency. What do you think happens when the least efficient part of the system is you?
I think Corridor Cast doesn't seems to be understand implication of the generative AI. For them it's tool to make more cheap contents. But the real issue is the future of the art as a whole. It cheapens the art into more generic stuff gives even more bad smelly stigma around it, basically driving away, stunting new generation of artists. It means more tech people creating generic shits, with no vision, creative direction and basically levels down for generations. There is no winning party of this AI generative arts unless it's cheap content makers, disillusioned ai art bros. It's stealing art and top of that even money they making, lively hood.
Podcasts are so much better without Jake
Fascinating I'm watching brilliant people through Internet, on my computer at home, with them talking about awesome innovations (in a recorded podcast - not just written down ideas) in movie and television video graphics with the aid of computers and now with state-of-the-art AI image generation..... what a time to be alive truly. So happy and thankful to be a fan of Corridor Digital.
So excited for this corridor project, killin me with the delayed release this week hahaha. It's gonna be so epic, I just know it.
I think this will be great for artists to feed their own images into a generator, create a model from their images and then animate using that model