Plant Library is only available for blender 3 and above. Which requires windows 10 or higher. So I do have a reason not to try it as I am not reinstalling an entirely new OS just to have a free addon. The cracked stuff works just fine.
I don't see a path for AI diffusion models to ultimately matter much or make money besides MJ and Adobe selling us subscriptions. Users are struggling to find a way to make AI relevant to their workflow(me included) and even with Control net, the opaqueness and randomness in AI makes it not that useful. We, artists are precise and giving away control to some random seed much too far down the pipeline is antithetical to what our clients hire us for. Another big problem for AI images(which I think is the only good thing) is that the outputs themselves are not copyrightable, making them worthless except for pitch decks, instagram posts and maybe low quality web ads. I remember the year uber and lyft got big(2011?) and since then we've all taken ride share all over, many times... yet uber still isn't profitable. I don't see a path for this stuff beside Adobe selling us another subscriptions service and in return we get what, clients and the general public value our effort less?
AI is a tool not a solution. Diffusive AIs can be super useful to create style guides (for artist teams) or Indie game devs to build art work really quickly. For eg: Diffusive AI can help an Indie dev save time trying to build artwork for game Icons.
@@libshastra I appreciate that take, but I just don't see game studios being comfortable with their IP being non-copyrightable. It's the very thing that makes them money. It's also the reason why in house assets are so important. I could see epic, or other studios like Disney have inhouse AI training on their IP which is copyrighted, but using random AI diffusion models is like downloading random unlicensed assets off of the open internet. Again, for one off assets, concepting, mood boards sure, but then it's just an advanced google search.
@@MrMadvillan Actually you can easily fine tune the models on your own assets and artstyle to make it use those instead. A properly fine tuned and pruned model shouldn't be able to generate assets based on style other than the one you've trained it on. If you look into the topic a little deeper you will find that there are hundreds of models trained to do just that. Using a carefully curated set of input art you can achieve much greater consistency. As an example you can fine tune a model on your own pictures and it will generate more pictures of specifically you with extremely high accuracy. You don't even need more than a doze of pictures. Just being able to use a few pictures of a character to then generate hundreds more in different, specified by you (using control net) poses would be invaluable to an indie developer. As an example this would greatly speed up the tedious process of creating 2D sprite animation assets. If you want more control, then you can simply draw a single sprite of the character and then use image to image to animate that specific character. If you want, you can take a look at "Rock Paper Scissors" and its making of videos by the Corridor crew. They have achieved enough consistency in their model to create a continuous animation! That is very far from random and unpredictable. Regarding the copyright it is a fair point although this is a very fresh field and many things may change. Would an intern tracing over the AI generated asset be copyrightable? It is not like there is a way to detect that and as such you would have to catch the studio red handed to deny them copyright. What if we replaced the intern with another machine? The line is still very much up in the air.
@Esphaeras Praestans I honestly don't think it will be long before heavy restrictions will be put on a AI training data. I can't imagine the government is going to risk a future where hyper real deepfakes exist totally unchecked. That won't stop company's from training their own in-house AI with their own copyrighted data sets, but at least it will relegate AI to a niche that we can hopefully learn to live with and perhaps even find useful.
Great list to evaluate from! Thank you for your efforts in sharing the knowledge! 👍
Thanks for sharing your knowledge! 👏👏👏
Thank you for sharing as always!
Great selection
Plant Library is only available for blender 3 and above. Which requires windows 10 or higher. So I do have a reason not to try it as I am not reinstalling an entirely new OS just to have a free addon. The cracked stuff works just fine.
Thanks m8!
Bruh..how many channels you have man. with same type of content. I swear every blender video I click I hear ur voice.
@ 3:55 - it's not "neutral" network - it's NEURAL network.
You must've thought the "t" was omitted because of a typo, right?
Or his AI auto spell correct did.
@@obsidianjane4413 Indeed! 😁
So you can now import a stick figure drawn in MS Paint and get a photoreal image...
I don't see a path for AI diffusion models to ultimately matter much or make money besides MJ and Adobe selling us subscriptions. Users are struggling to find a way to make AI relevant to their workflow(me included) and even with Control net, the opaqueness and randomness in AI makes it not that useful. We, artists are precise and giving away control to some random seed much too far down the pipeline is antithetical to what our clients hire us for. Another big problem for AI images(which I think is the only good thing) is that the outputs themselves are not copyrightable, making them worthless except for pitch decks, instagram posts and maybe low quality web ads. I remember the year uber and lyft got big(2011?) and since then we've all taken ride share all over, many times... yet uber still isn't profitable. I don't see a path for this stuff beside Adobe selling us another subscriptions service and in return we get what, clients and the general public value our effort less?
AI is a tool not a solution. Diffusive AIs can be super useful to create style guides (for artist teams) or Indie game devs to build art work really quickly. For eg: Diffusive AI can help an Indie dev save time trying to build artwork for game Icons.
@@libshastra I appreciate that take, but I just don't see game studios being comfortable with their IP being non-copyrightable. It's the very thing that makes them money. It's also the reason why in house assets are so important. I could see epic, or other studios like Disney have inhouse AI training on their IP which is copyrighted, but using random AI diffusion models is like downloading random unlicensed assets off of the open internet. Again, for one off assets, concepting, mood boards sure, but then it's just an advanced google search.
@@MrMadvillan Actually you can easily fine tune the models on your own assets and artstyle to make it use those instead. A properly fine tuned and pruned model shouldn't be able to generate assets based on style other than the one you've trained it on. If you look into the topic a little deeper you will find that there are hundreds of models trained to do just that. Using a carefully curated set of input art you can achieve much greater consistency.
As an example you can fine tune a model on your own pictures and it will generate more pictures of specifically you with extremely high accuracy. You don't even need more than a doze of pictures.
Just being able to use a few pictures of a character to then generate hundreds more in different, specified by you (using control net) poses would be invaluable to an indie developer. As an example this would greatly speed up the tedious process of creating 2D sprite animation assets. If you want more control, then you can simply draw a single sprite of the character and then use image to image to animate that specific character.
If you want, you can take a look at "Rock Paper Scissors" and its making of videos by the Corridor crew. They have achieved enough consistency in their model to create a continuous animation! That is very far from random and unpredictable.
Regarding the copyright it is a fair point although this is a very fresh field and many things may change. Would an intern tracing over the AI generated asset be copyrightable? It is not like there is a way to detect that and as such you would have to catch the studio red handed to deny them copyright. What if we replaced the intern with another machine? The line is still very much up in the air.
@Esphaeras Praestans do you actually do 3d professionaly?
@Esphaeras Praestans I honestly don't think it will be long before heavy restrictions will be put on a AI training data. I can't imagine the government is going to risk a future where hyper real deepfakes exist totally unchecked. That won't stop company's from training their own in-house AI with their own copyrighted data sets, but at least it will relegate AI to a niche that we can hopefully learn to live with and perhaps even find useful.
Can you share my project ??
These are not free btw lol.
Its the scam they play. Say free. Which it is sort of as a trial. But it is not free.
I guess freedom of speech means freedom to lie.
The model yes the equipment no