The value of human made will rise because the panic in many people. Less and less pursue art -> value of those who do pursue art increases. The crap can hit the fan at any moment once the proper copyright laws will come into place. Secondly generating stuff and especially the AI learning process IS SUPER EXPENSIVE! Rich companies like nVidia are dumping money like there is no tomorrow, but they know and those who work on the AI field know this is a bubble. We would need to find a new powerful yet cheap energy source (like the fusion reactor if that ever will happen) to justify the costs.
Current copyright law is already too restrictive. We will likely see it loosened before it's ever brought to bear on transformative works like AI produces.
AI generation is extremely cheap, GPT-4 is 99% cheaper compared to GPT-3, and the best image generation model is free, open source, and can run on any recent hardware locally. And this trend will continue with newer and better models. There is no such thing as an AI bubble. Most will not even care for human-made things, we just want the better and cheaper things.
Keep coping. AI is no bubble. Its the future. The Quality Jump from version 3 to 4 is big. A few more Versions down the line and this will be perfectly usable. And this goes for a lot of AI tools.
@@YankeeDoodle-hr9dvYeah, that is why people have every stuff made in china. Because they give a damn about where stuff comes from and under which conditions it was made and how the workers got paid. No, this world is about consume and in a few years you can't say, if AI or human made it and nobody will care.
you could have blocked out most of these within 30 minutes and been well on your way to fleshing out a model with good topology. All of these will need significant retopology work. Some of them need to be completely overhauled. That hard surface model of the engine is unusable. Would not work for any application. It looked like a poorly done LIDAR scan with 1000s of triangles. Not sure how this AI model generator is supposed to streamline the 3D modeling pipeline when the results are 90% garbage. you will use all 8 tokens or whatever immediately with no promise of a usable model.
Even the textures that are generated not good for reusing. Those are not neutral textures, but shaded, with all the lights applied to them, means you can't use those textures without redrawing everything to make them unlit, unshaded.
Even though this is a sponsor video, personally this tool helps me speed up working with concepts at the moment, and when open source local 3D models of the same level or higher come out, I will switch to them, but for now there are none. It's like Midjourney for 3D, until SD for 3D appears, we'll have to buy services like this.
So it exports with the rig? Can it do face rigging or creating a mesh that is capable of face rigging? I notice these programs have a solid face so i have to put in alot of work to be able to animate the face
Rodin (another ai to 3d service, did a video about them as well) can kind of do that, except it takes front/side/top views as input. They now added a new feature that allows you to upload a simple mesh which is used to define the overall shape of the 3d model. That gives you plenty of control over the generation.
15 credits for mesh generation + 10 credits for texturing task per mesh. At least give us textured variations of those generated ones so we can see the result before trying to milk my wallet. Such an obvious scam made out of AI trend.
I guess so. Modeling will probably be less in demand than even in the near future. Although it has been outsourced and made cheap long ago, so this isn't such a huge deal imho. Instead of being worried, focus on skills that can't be automated yet (such as being a unique human) and using the new tools to make higher level stuff.
I remember a few years ago when I said this would happen and we need to beachful before this stuff starts replacing people. I was called stupid and labeled a luddite. Well at least I get to say I told you so to a bunch of people. Not that it will make me fell much better after spending the last 30 years learning this crap.
Garbage tool made for those that have no skills... and, for some reason, don't want to develop any. If you think this is cool or that the models are good/usable its because you have NO IDEA what an appealing and good model is.
@@CGDive so there is a guide you can check out... actually, there are so many guides you can check out about 3D modeling, they're called tutorials, and if you spend some time with them, you might actually learn something and develop a skill that you can use in the future. You didn't made that, you just pressed a button. Where is the fun of having to create something cool? Why on earth would you waste your time with this? Sorry, but its actually garbage, you take out the part where you learn new things, and get nothing in return... And it's not virtue signalling, it's just the facts my guy
@@botondszilagyi3713 "You didn't made that, you just pressed a button." True. "Where is the fun of having to create something cool?" Not everyone is having fun the same way. If you have fun 3d modeling, model. But for other people the fun is for example in telling a story, even if they did not create every single polygon. That's the skill they want to develop. These people often use model libraries and in the future they will probably use AI generated 3D models. I don't think you are trying to virtue signal. But I think you are measuring things by your own yard-stick, and making judgements that are not true for everyone.
@@CGDive If they do not list any information about the data source then it is most likely web scraped content. I would be careful about promoting such products. Especially as this is a CG channel that is directly related to people whos content is violated
@@dreww2647 It depends on the licence. Many models that are free to download require attribution and it is widely known that web scrapers ignore this. You can't give attribution when you don't know what the AI is using to generate your prompt
@lapissea1190 Legally, you don't need to give attribution for training data. Generative AI is a probabilistic error correction system. Training only defines trends for the system to favor. It's like painting a study of Michelangelo's work in order to learn how to paint in a sculptural style. Totally transformative by any definition of the word. The output can't be copyrighted, though, because it was machine produced.
*Check out Meshy AI*
app.meshy.ai/discover?via=cgdive
The value of human made will rise because the panic in many people.
Less and less pursue art -> value of those who do pursue art increases.
The crap can hit the fan at any moment once the proper copyright laws will come into place.
Secondly generating stuff and especially the AI learning process IS SUPER EXPENSIVE!
Rich companies like nVidia are dumping money like there is no tomorrow, but they know and those who work on the AI field know this is a bubble.
We would need to find a new powerful yet cheap energy source (like the fusion reactor if that ever will happen) to justify the costs.
Current copyright law is already too restrictive. We will likely see it loosened before it's ever brought to bear on transformative works like AI produces.
AI generation is extremely cheap, GPT-4 is 99% cheaper compared to GPT-3, and the best image generation model is free, open source, and can run on any recent hardware locally.
And this trend will continue with newer and better models.
There is no such thing as an AI bubble.
Most will not even care for human-made things, we just want the better and cheaper things.
@@OnigoroshiZero Speak for yourself. I can assure you that a lot of people value human made things (and I'm not just talking about art)
Keep coping. AI is no bubble. Its the future. The Quality Jump from version 3 to 4 is big. A few more Versions down the line and this will be perfectly usable. And this goes for a lot of AI tools.
@@YankeeDoodle-hr9dvYeah, that is why people have every stuff made in china. Because they give a damn about where stuff comes from and under which conditions it was made and how the workers got paid. No, this world is about consume and in a few years you can't say, if AI or human made it and nobody will care.
Pretty cool in my opinion, definitely a great starting point for a model
Would you recommend meshy for a mobile game? I dont need highly detailed characters since it wont be for pc
I am not much of a recommender, sorry. You see what it does. If it seems to cover what you need, give it a go.
you could have blocked out most of these within 30 minutes and been well on your way to fleshing out a model with good topology. All of these will need significant retopology work. Some of them need to be completely overhauled. That hard surface model of the engine is unusable. Would not work for any application. It looked like a poorly done LIDAR scan with 1000s of triangles. Not sure how this AI model generator is supposed to streamline the 3D modeling pipeline when the results are 90% garbage. you will use all 8 tokens or whatever immediately with no promise of a usable model.
Even the textures that are generated not good for reusing. Those are not neutral textures, but shaded, with all the lights applied to them, means you can't use those textures without redrawing everything to make them unlit, unshaded.
Even though this is a sponsor video, personally this tool helps me speed up working with concepts at the moment, and when open source local 3D models of the same level or higher come out, I will switch to them, but for now there are none.
It's like Midjourney for 3D, until SD for 3D appears, we'll have to buy services like this.
So it exports with the rig? Can it do face rigging or creating a mesh that is capable of face rigging? I notice these programs have a solid face so i have to put in alot of work to be able to animate the face
I would love it when AI can use a character turnaround to create a model.
Rodin (another ai to 3d service, did a video about them as well) can kind of do that, except it takes front/side/top views as input. They now added a new feature that allows you to upload a simple mesh which is used to define the overall shape of the 3d model. That gives you plenty of control over the generation.
15 credits for mesh generation + 10 credits for texturing task per mesh. At least give us textured variations of those generated ones so we can see the result before trying to milk my wallet. Such an obvious scam made out of AI trend.
HAHAHAHAA brother said generate the rig.
haha, it's a habit :P
Should 3d modellers be worried?
in a near futur yes absolutely
I guess so. Modeling will probably be less in demand than even in the near future. Although it has been outsourced and made cheap long ago, so this isn't such a huge deal imho.
Instead of being worried, focus on skills that can't be automated yet (such as being a unique human) and using the new tools to make higher level stuff.
What are some skills that will be automated at last , what's yours opinion@@CGDive
I remember a few years ago when I said this would happen and we need to beachful before this stuff starts replacing people. I was called stupid and labeled a luddite. Well at least I get to say I told you so to a bunch of people. Not that it will make me fell much better after spending the last 30 years learning this crap.
@@nimay321server repair.
50k triangles are too much. This thing needs much lower poly count options like 500, 1k and 3k.
You could do the retopo in Blender.
That's where the human comes into play 😂😂 good hold manual retopo and uv layouts 😂😂
Garbage tool made for those that have no skills... and, for some reason, don't want to develop any.
If you think this is cool or that the models are good/usable its because you have NO IDEA what an appealing and good model is.
I see this video is a virtue signaling paradise. Keep at it, it's making the video more popular.
@@CGDive so there is a guide you can check out... actually, there are so many guides you can check out about 3D modeling, they're called tutorials, and if you spend some time with them, you might actually learn something and develop a skill that you can use in the future. You didn't made that, you just pressed a button. Where is the fun of having to create something cool? Why on earth would you waste your time with this? Sorry, but its actually garbage, you take out the part where you learn new things, and get nothing in return... And it's not virtue signalling, it's just the facts my guy
@@botondszilagyi3713
"You didn't made that, you just pressed a button."
True.
"Where is the fun of having to create something cool?"
Not everyone is having fun the same way. If you have fun 3d modeling, model. But for other people the fun is for example in telling a story, even if they did not create every single polygon. That's the skill they want to develop. These people often use model libraries and in the future they will probably use AI generated 3D models.
I don't think you are trying to virtue signal. But I think you are measuring things by your own yard-stick, and making judgements that are not true for everyone.
Is this model trained on ethically trained data or is this laundering of copyright again
I can't say for sure. Most AI companies will say they are ethical but it's almost impossible to know what goes on behind closed doors.
@@CGDive If they do not list any information about the data source then it is most likely web scraped content. I would be careful about promoting such products. Especially as this is a CG channel that is directly related to people whos content is violated
Yes, all data is fair use for training. No AI work is copyrightable. Hope this helps.
@@dreww2647 It depends on the licence. Many models that are free to download require attribution and it is widely known that web scrapers ignore this. You can't give attribution when you don't know what the AI is using to generate your prompt
@lapissea1190 Legally, you don't need to give attribution for training data. Generative AI is a probabilistic error correction system. Training only defines trends for the system to favor. It's like painting a study of Michelangelo's work in order to learn how to paint in a sculptural style.
Totally transformative by any definition of the word. The output can't be copyrighted, though, because it was machine produced.
Jus say no to AI..
Elaborate.
Yo pude hacer modelos realistas, literalmente como si fueran humanos reales, solo tenes que ajustar el prompt
Good to know!
Awesome A.i. software generation. Just playing with it just now and it generates models that I could build off of, improve and sell as assets 0.o
it`s disgustivg when devs overcharge their NOT FINISHED PRODUCT before they even fix all major bugs.
THE RESULTS ARE NOT DESERVE SO MUCH MONEY!
It's 50 cents per model. I can understand not liking the service but as far as price goes, I am not sure how it can get cheaper.