Ok 1. Mootion, I am going to check out your platform because of this video. Keep sponsoring this dude. 2. I’d love to see a video where you combine a lot of the elements you’ve talked about together to make a music video starting from lyrics. I’ve gotten comfortable generating the audio with Suno, but am still trying to figure out what platforms will provide the best results for creating consistent characters, turning those characters into video, adding lip synch, adding animated lyrics, etc. again, going to check out Mootion for this.
This needs to be discussed; Ai training is not Stealing nor is it copying. That is why every lawsuit so far has been either thrown out or dismissed. Ai training is literally just taking "inspiration" from those things. You know what every single person does. If that is grounds for a lawsuit then everyone can get sued. We need to fight the mass disinformation about Ai training with critical thinking and common sense in objective based reality.
It's a touchy subject for the time being, but I'm in agreement with you, and I tend to think within the next few years, it'll all get sorted out an normalized. There will still be people who steadfastly refuse to use the tools, just as there are still artists who choose not to work digitally...but, for the majority of work, AI is simply going to be as common as using Photoshop. If you ever go back and watch some news reports from the release of Photoshop v.1, you'll see a lot of the same arguments. It's just a part of the progression of technology.
Holy crap!!!!! I,.paused your vid for a few hours, maxed out the Moverse demo, and walked away with 22 minutes of my movie competed! Amazing... Ai-vid is making years of technology progress pass by in hours. Thank you Theoretically Media for forging the path!!!!! :)
I think the thing these blogs need to remind everyone is that there is no law whatsoever that can keep an AI company from doing things like this and a TOS (Term of Service) is not a legally binding contact. Judges typically chuck them to the wind in the courtroom. At the most, should you break a TOS, they can cancel your account, that's it.
Oh, for sure. Again: Not A Lawyer, but the whole training data thing is a done deal. Even if the US tries to stamp out any kind of training, you've still got an entire globe to which those laws don't apply...And, well-- the internet doesn't care about silly little lines on a map. I know Japan has already OK'd all training data...so, y'know- maybe we'll just be using .jp at the end of web addresses soon.
It appears you're attempting to cope with this all with false claims. A TOS IS a legally binding contract. Courts typically do uphold TOS agreements (as long as they aren't unconscionable). 'AI bros' don't take into consideration data protection laws, consumer protection laws, and many more. _ This statement oversimplifies the legal landscape and heavily underestimates the enforceability of TOS agreements as well as the regulatory environment for AI companies.
If a company uses UA-cam videos by means of training their machines, UA-cam has all rights to take legal action. Courts USUALLY uphold TOS agreements (if they're clear / if there is a general user consensus). The company actually can be sued due to something called 'breach of contract' and UA-cam can seek an injunction. You can argue all you want that publicly available = fair game, but that doesn't align at all with the law. UA-cam videos are protected by copyright and-- unless it is labeled as fair use --scraping these videos would be heavily scrutinized in court. _ I also urge you to take into consideration the Computer Fraud and Abuse Act (CFAA)
It's not bad, right? Seriously makes things a TON easier! I've been playing around with it by loading in a lot of Midjourney images for inspired prompts, and I'm actually kind of impressed with how the generations are turning out!
I know I was being a little jokey with the Refield bit, but Resona is pretty good. The fact that you can have contextual sound effects generated in a video? That's awesome.
They’ve been around for a bit, and were really early in the “Text prompt a 3D character to do an action or motion” - but, also Discord based, so I think there was a good amount of people who never found them. With Storyteller, it looks like they’re starting to move (ahem) into a 2.0 era. And I think once they get the platform solidified, they’ll be a powerhouse.
Hi Tim, thanks for the video and for the prompt generator! I'm pretty bad at prompting so I need all the help I can get. 🤣 Love that you are sharing so much information with everyone!
I was going to, but at this point I've done so many videos on Kling. And, y'know everyone else is doing videos on it-- I figured I'd just go in a different direction.
The snake eating its tail. Or...actually, rather the analog video that is copied and then copied again. Although, you can get some cool experimental results when you do that!
that's false. it's called synthetic data and there's nothing wrong in it. if we used your logic, humans would become dumber when using data or content made by other humans.
I think that the end result of this training is going to be some sort of standardized licensing specifically for video and music used to train AI. Licensing already exists for other use types, but this is really different because of how training works. The works aren’t being duplicated in whole or utilized in any recognizable way as the end product. The user is generating something original that’s based on what AI learned. It’s going to be interesting, but necessary for sure. I think the outcome could be good for both the original artists and the end users. The users will be able to get results based on training from professional quality works while the original artists get compensation - hopefully fair, although it might not be after their publishers get finished pulling out their share (what else is new). Maybe there will be a way for artists to go to AI deveolpers directly and cut out that third party loss. Personally, I’m not so sure the fair use doctrine doesn’t apply. It might be decided that it does. Either way, lots of things will change.
so....gen 3 doesnt have image to video..only text to video??? so i cant use midjourney images to animate using gen 3?? i can only use gen 2 for image to video. this sucks !
The birdwatcher in me finds that AI videos are not too fussy at mixing and mashing the real world. Those "toucans" are a hornbill blend (almost accurate Great Hornbill) in the Motion Storyteller clip and that tropical beach scene sounds like it has been invaded by Herring Gulls seeking warmer climes. Its like those Kookaburra calls you hear in African Jungles -thanks Tarzan
I love everything about this comment. Everything. Sidenote: the house across the street from me has a Bald Eagle nest. They've been returning for the past few years now, so it looks like we've got some long terms neighbors. It's been pretty cool watching them zoom over while I'm out walking the dog!
If you make AI while taking other peoples work, regardless of its current legality or not, you are simply signing yourself up for future problems. Beware if you give these companies our money long termed. Edit: Also, thanks for the video good sir!
Appreciated! And yeah-- it's going to be sticky for awhile. I liken it to sampling/DJing in Hip Hop. We're really early in, and those needle drops had a LOT of legality issues at the dawn of rap-- but, obviously that was all worked out, and now we have a global music genre. It'll work out eventually.
@@TheoreticallyMedia True and good analogy however & I am sure you know this 100% but those samples we use when making beats...they are using older samples/royalty free/non copywritten or purchased the rights to...They are not just taking samples scraped en mass and releasing them without compensation or legal rights to (&those who did got nuked in the industry). It worked out now but thats kinda the whole point here: the way it worked out in rap was the producers of beats capitulated due to them not wanting to be fined massively.
Humans learn to make art by examining the art of those who have come before. AI is no different, only a lot faster. Whether its still images or UA-cam videos, nothing is being copied directly, it is a mishmash of billions of pieces of information/images. It is the very LEGAL definition of Fair Use in that it is transformative. I am very disappointed in all the AI companies who are rolling over when this should be a slam dunk defense - assuming you don't get some 100 year old judge who can't even open his email.
How can you legally and logically say that giving access to a video, or whatever, to an AI system is any different from giving access to an human? Watching is watching.
The counterargument is the speed at which AI processes all that information-- but, y'know-- I'm obviously on your side here. That, and provided the output isn't a 1:1 copy as well.
I find it hilarious (extremely hypocritical?) that people are up in arms about scraping thousands of UA-cam videos and transcripts, when they defended scraping billions of artist's works (I love the guy, but Marques Brownlee was a huge crybaby about this whole non-event). I don't see this as any different and think anyone should be able to train on all of this data as long as they are not reproducing exact replicas (which I have never seen happen). Even the NYTimes suit against AI was basically the Times prompting the AI very precisely for excerpts of their full text articles and then being outraged that it sort of almost created excerpts from their full text articles. Talk about leading the witness or entrapment! Anyway, that's my Hot Take.
I'm not saying my hot take! But I'm not saying I don't disagree with your take. But I'm not, NOT saying that either! haha...I think I covered myself with that confusing comment.
I liked the concept and visited the site storyteller but it will not take regular photos and the images the program uses are not in keeping with the script at all. For instanne my script had a nine year old girl walking to school meeting a fairy on the way and talking to her and all the site had were images of a fairy. I could not upload any photos.
Yeah, it for sure takes creative liberties currently. I’d say your best results will occur when you give it broad story ideas at let it co-create with you. Technically, you could go through and edit everything back to the way you wanted your story as well?
There's a few popping up in Asia as well-- Japan, for one, allows training. So-- yeah, to be totally honest: the US needs to get over it, or it'll be left behind in the AI Video race.
if someone used a tiny piece of something i created and they put it into a Montage.....Id be happy - also not aware of anyone ever complain about Montage's
Wow, that ex Runway employee must really hate those guys to leak that 😂. In my dealings with them they’ve just always reaked of typical wrong headed start up bros steering the wheels there and all their Gen-2 code deployments for new features were always non QA’d fubar shit shows where people lost their work and they just shrugged and chugged along taking folks money.
WIthout data there would be no AI, so if we deal with this issue then basically AI is done here in the west. Meanwhile countries like China can do whatever they want and continue to build more and more powerful models of all kinds.
So i could go onto any hire freelance site (fiverr, upwork, freelancer, people per hour) and pay them to copy a pixar film style , look, finish and thsts ok. But if i pay for a computer to do it.... thaaaaaaats baaaaaaad.
Some stuff was redacted, but make sure you check out the tabs on the bottom. It might look empty, but once you switch tabs, you'll see a lot more info.
Ok 1. Mootion, I am going to check out your platform because of this video. Keep sponsoring this dude. 2. I’d love to see a video where you combine a lot of the elements you’ve talked about together to make a music video starting from lyrics. I’ve gotten comfortable generating the audio with Suno, but am still trying to figure out what platforms will provide the best results for creating consistent characters, turning those characters into video, adding lip synch, adding animated lyrics, etc. again, going to check out Mootion for this.
This needs to be discussed; Ai training is not Stealing nor is it copying. That is why every lawsuit so far has been either thrown out or dismissed. Ai training is literally just taking "inspiration" from those things. You know what every single person does. If that is grounds for a lawsuit then everyone can get sued. We need to fight the mass disinformation about Ai training with critical thinking and common sense in objective based reality.
It's a touchy subject for the time being, but I'm in agreement with you, and I tend to think within the next few years, it'll all get sorted out an normalized. There will still be people who steadfastly refuse to use the tools, just as there are still artists who choose not to work digitally...but, for the majority of work, AI is simply going to be as common as using Photoshop.
If you ever go back and watch some news reports from the release of Photoshop v.1, you'll see a lot of the same arguments. It's just a part of the progression of technology.
So if its not theft AI programmers can copy Chat gpt4 - for training purposes...
3:43
Could have VERY major model training advantages over all competitors! 😲
Holy crap!!!!! I,.paused your vid for a few hours, maxed out the Moverse demo, and walked away with 22 minutes of my movie competed!
Amazing... Ai-vid is making years of technology progress pass by in hours.
Thank you Theoretically Media for forging the path!!!!!
:)
I'll be back next month to continue:
ua-cam.com/video/uESEAt0GtrI/v-deo.htmlsi=nHLGVrwla3dSQpCG
I think the thing these blogs need to remind everyone is that there is no law whatsoever that can keep an AI company from doing things like this and a TOS (Term of Service) is not a legally binding contact. Judges typically chuck them to the wind in the courtroom. At the most, should you break a TOS, they can cancel your account, that's it.
Oh, for sure. Again: Not A Lawyer, but the whole training data thing is a done deal. Even if the US tries to stamp out any kind of training, you've still got an entire globe to which those laws don't apply...And, well-- the internet doesn't care about silly little lines on a map.
I know Japan has already OK'd all training data...so, y'know- maybe we'll just be using .jp at the end of web addresses soon.
Exactly, there's nothing illegal about downloading UA-cam videos
It appears you're attempting to cope with this all with false claims. A TOS IS a legally binding contract. Courts typically do uphold TOS agreements (as long as they aren't unconscionable).
'AI bros' don't take into consideration data protection laws, consumer protection laws, and many more.
_
This statement oversimplifies the legal landscape and heavily underestimates the enforceability of TOS agreements as well as the regulatory environment for AI companies.
If a company uses UA-cam videos by means of training their machines, UA-cam has all rights to take legal action. Courts USUALLY uphold TOS agreements (if they're clear / if there is a general user consensus).
The company actually can be sued due to something called 'breach of contract' and UA-cam can seek an injunction.
You can argue all you want that publicly available = fair game, but that doesn't align at all with the law. UA-cam videos are protected by copyright and-- unless it is labeled as fair use --scraping these videos would be heavily scrutinized in court.
_
I also urge you to take into consideration the Computer Fraud and Abuse Act (CFAA)
@@jumpy2783 nothing it going to happen. Cope
Big THANK YOU for the Gen-3 prompt Generator!!
It's not bad, right? Seriously makes things a TON easier! I've been playing around with it by loading in a lot of Midjourney images for inspired prompts, and I'm actually kind of impressed with how the generations are turning out!
Kling also released their AI Video Generator globally!
Resona is a pretty good concept, might as well use it for my AI videos
I know I was being a little jokey with the Refield bit, but Resona is pretty good. The fact that you can have contextual sound effects generated in a video? That's awesome.
Super interesting, will try right now for our channel. Haven't heard of Mootion before, so thank you 🙂
They’ve been around for a bit, and were really early in the “Text prompt a 3D character to do an action or motion” - but, also Discord based, so I think there was a good amount of people who never found them.
With Storyteller, it looks like they’re starting to move (ahem) into a 2.0 era. And I think once they get the platform solidified, they’ll be a powerhouse.
You nailed it! Runway ml's missing link is training on Theoretically Media like we do! Maybe in the next update?
Haha, they have an open invitation! Happy to send them over every video I've ever done!
Hi Tim, thanks for the video and for the prompt generator! I'm pretty bad at prompting so I need all the help I can get. 🤣 Love that you are sharing so much information with everyone!
I’m surprised you didn’t talk about kling having world access now
I was going to, but at this point I've done so many videos on Kling. And, y'know everyone else is doing videos on it-- I figured I'd just go in a different direction.
@@TheoreticallyMedia no we always want a Tim perspective and new video
I used Kling it’s not that impressive to call it the Sora killer.
You can't train an LLM on content that came from an LLM. It creates a recursive detail loop that will destroy the output fidelity.
The snake eating its tail. Or...actually, rather the analog video that is copied and then copied again.
Although, you can get some cool experimental results when you do that!
that's false. it's called synthetic data and there's nothing wrong in it. if we used your logic, humans would become dumber when using data or content made by other humans.
Thank you for all your hard work and for taking your time to share with us. You're pretty great on camera.
Appreciate that! Haha, personally, I hate watching myself-- but I think that's true for most of us! (Editor me also hates "on camera" me!)
nice thank you for the runway custom GPT prompter
Excellent to hear!! Hope you're getting some good result from it!
For everyone reading this I hope you all have a great day, I’m off to see Deadpool with the wife. ps ACCELERATE
Ahhhh! Can't WAIT to see that! Might have to play hooky sometime next week to sneak in a screening!
BTW, i finally got access to Kling but haven't used it yet
Excellent to hear! You’ll have a lot of fun once you get a chance. Currently, the best Image to Video around, in my opinion.
Balloon over turkey is not super scifi environment... It's the landscape of Kapadokia, real 😅
Seriously?! Man-- I gotta go there at some point! That's nuts!
I think that the end result of this training is going to be some sort of standardized licensing specifically for video and music used to train AI. Licensing already exists for other use types, but this is really different because of how training works. The works aren’t being duplicated in whole or utilized in any recognizable way as the end product. The user is generating something original that’s based on what AI learned. It’s going to be interesting, but necessary for sure.
I think the outcome could be good for both the original artists and the end users. The users will be able to get results based on training from professional quality works while the original artists get compensation - hopefully fair, although it might not be after their publishers get finished pulling out their share (what else is new). Maybe there will be a way for artists to go to AI deveolpers directly and cut out that third party loss.
Personally, I’m not so sure the fair use doctrine doesn’t apply. It might be decided that it does. Either way, lots of things will change.
Great info here. Can you do a video on how you created the Custom Gpt from the Spreadsheet data?
That's a great idea! Will do!
@@TheoreticallyMedia Thanks!
I want to watch your new vid, but I'm too busy implementing what you taught me already!
:)
Thank you!!!
Haha, this one is a little on the "news" side-- so, you can watch it in the background!
Alright! I'm back :)
Great stuff Tim! You rule.
Thanks so much!! As always!!
so....gen 3 doesnt have image to video..only text to video??? so i cant use midjourney images to animate using gen 3?? i can only use gen 2 for image to video. this sucks !
Yup. The I2V is the big thing holding Gen-3 back right now, in my opinion at least.
Still a good model, just hampered by the lack of Image Input.
very interesting news, thank you for the video
1000%! Thank you for checking it out!
The birdwatcher in me finds that AI videos are not too fussy at mixing and mashing the real world. Those "toucans" are a hornbill blend (almost accurate Great Hornbill) in the Motion Storyteller clip and that tropical beach scene sounds like it has been invaded by Herring Gulls seeking warmer climes. Its like those Kookaburra calls you hear in African Jungles -thanks Tarzan
I love everything about this comment. Everything.
Sidenote: the house across the street from me has a Bald Eagle nest. They've been returning for the past few years now, so it looks like we've got some long terms neighbors. It's been pretty cool watching them zoom over while I'm out walking the dog!
Thanks for the video.
You bet
Tim is the best
Makin' me blush over here!
If you make AI while taking other peoples work, regardless of its current legality or not, you are simply signing yourself up for future problems. Beware if you give these companies our money long termed.
Edit: Also, thanks for the video good sir!
Appreciated!
And yeah-- it's going to be sticky for awhile. I liken it to sampling/DJing in Hip Hop. We're really early in, and those needle drops had a LOT of legality issues at the dawn of rap-- but, obviously that was all worked out, and now we have a global music genre.
It'll work out eventually.
@@TheoreticallyMedia True and good analogy however & I am sure you know this 100% but those samples we use when making beats...they are using older samples/royalty free/non copywritten or purchased the rights to...They are not just taking samples scraped en mass and releasing them without compensation or legal rights to (&those who did got nuked in the industry).
It worked out now but thats kinda the whole point here: the way it worked out in rap was the producers of beats capitulated due to them not wanting to be fined massively.
Thanks for this video, nice coverage
Heck yeah Tim u rock the fact making us a gpt rocks n can u do a video run some of these gems in runway 3 lol n ur jokes always best lol ❤
Bless!! I hope the GPT is working out for you! It can be hit or miss, but I tend to think hits more often than not!
Humans learn to make art by examining the art of those who have come before. AI is no different, only a lot faster. Whether its still images or UA-cam videos, nothing is being copied directly, it is a mishmash of billions of pieces of information/images. It is the very LEGAL definition of Fair Use in that it is transformative. I am very disappointed in all the AI companies who are rolling over when this should be a slam dunk defense - assuming you don't get some 100 year old judge who can't even open his email.
Thank you 🙂
Cool
Pretty neat stuff this week!
How can you legally and logically say that giving access to a video, or whatever, to an AI system is any different from giving access to an human? Watching is watching.
The counterargument is the speed at which AI processes all that information-- but, y'know-- I'm obviously on your side here. That, and provided the output isn't a 1:1 copy as well.
I find it hilarious (extremely hypocritical?) that people are up in arms about scraping thousands of UA-cam videos and transcripts, when they defended scraping billions of artist's works (I love the guy, but Marques Brownlee was a huge crybaby about this whole non-event). I don't see this as any different and think anyone should be able to train on all of this data as long as they are not reproducing exact replicas (which I have never seen happen).
Even the NYTimes suit against AI was basically the Times prompting the AI very precisely for excerpts of their full text articles and then being outraged that it sort of almost created excerpts from their full text articles. Talk about leading the witness or entrapment! Anyway, that's my Hot Take.
I'm not saying my hot take! But I'm not saying I don't disagree with your take. But I'm not, NOT saying that either!
haha...I think I covered myself with that confusing comment.
@@TheoreticallyMedia perfectly understood! Wink wink, lol!
I liked the concept and visited the site storyteller but it will not take regular photos and the images the program uses are not in keeping with the script at all. For instanne my script had a nine year old girl walking to school meeting a fairy on the way and talking to her and all the site had were images of a fairy. I could not upload any photos.
Yeah, it for sure takes creative liberties currently. I’d say your best results will occur when you give it broad story ideas at let it co-create with you.
Technically, you could go through and edit everything back to the way you wanted your story as well?
This is also why KeLing is gonna end up on top because the PRC couldn’t care less about American copyright laws😂.
There's a few popping up in Asia as well-- Japan, for one, allows training.
So-- yeah, to be totally honest: the US needs to get over it, or it'll be left behind in the AI Video race.
Would you recommend unlisting videos that didnt perform well and its obvious why?
if someone used a tiny piece of something i created and they put it into a Montage.....Id be happy - also not aware of anyone ever complain about Montage's
Wow, that ex Runway employee must really hate those guys to leak that 😂. In my dealings with them they’ve just always reaked of typical wrong headed start up bros steering the wheels there and all their Gen-2 code deployments for new features were always non QA’d fubar shit shows where people lost their work and they just shrugged and chugged along taking folks money.
Jesus Christ loves you all
WIthout data there would be no AI, so if we deal with this issue then basically AI is done here in the west. Meanwhile countries like China can do whatever they want and continue to build more and more powerful models of all kinds.
So i could go onto any hire freelance site (fiverr, upwork, freelancer, people per hour) and pay them to copy a pixar film style , look, finish and thsts ok.
But if i pay for a computer to do it.... thaaaaaaats baaaaaaad.
Just an empty doc - looks like all the entries have been deleted. Waste of time.
Some stuff was redacted, but make sure you check out the tabs on the bottom. It might look empty, but once you switch tabs, you'll see a lot more info.
This is really bad. Because this kind of copyright scale is big enough to cause World War 3 😅
Very interesting video. Thank you.
Thanks so much for checking it out!!