dreamcutter posted at 4:19 PM Sun, 26 November 2023 - #4478201

This AI 3d project seems to be aiming for that, applying animation to a fbx figure based on video/image samples .  https://plask.ai/  

However while many of the objectives you describe like text UI administration of Poser (Py script)  can already and has been done WITHOUT AI, the reliance of an AI webservice to the Poser software application will be a costly dependency and likely convert the use license to subscription or pay per use. Because it woudl so fundamentally change the direction and technical resources behind the development of Poser,  IMO it woudl make more sense for Bondware to license out the Poser API to a 3D AI art front end if the technology required it. 

I tried some of the suggested tools, that masterpieceX output is not render worthy... not close.  Maybe background characters for some amature low poly retro game, but not art quality renders,  Seems like embracing AI at the moment would be be foolhardy and reactionary to the hype and buzz... Poser is not NVIDIA.  I do not think Poser NEEDS AI integration to succeed... perhaps the depreciated FUSION capability could be replaces with OpenUSD & Omniverse support.  That would enable more than just AI in the pipeline.

I fully agree with you, the tools at the moment are not useful in real production. But looking at tendencies in these tools and what they are talking about, I think it is pretty obvious what they are aiming towards, so think ahead 2-5 years from now, if they can already achieve these things now. 

If you take the last video I linked. At the beginning of it, simply being able to take a 2d image, upload it to a website, and turn it into 3d in a few hours is pretty crazy. Same with how he rigs it and animates it.

As technology enhances there is a good chance that you will run this locally like you will do with SD. I think you can look at this as with videos, in the beginning (few month ago) you needed very powerful GPUs, now they are talking about how to render videos on less than 8GB of VRAM.

So I don't think one should judge these things on what we see today, because again they are in the very early stages of development. And sure Poser is not Nvidia or Microsoft, but a lot of these tools are not created by these huge companies. A lot of them make use of open-source stuff like SD, which is also in the process of creating a huge 3D model like they have for the 2D images, so it will be interesting to see that when it is done because I think that is when we will start to see real huge improvements in the 3D area.

And even if Poser doesn't have direct AI support, it doesn't change that people will simply start to use external 3D generators to create content rather than buying them, when these tools become good enough. If you can upload an image of a leather sofa or use text to 3d in SD to generate a high quality fully textured 3D object then you won't buy it at DAZ or Renderosity. 

No different than you can download free 3d objects from websites now, the only issue with them is that they often have a lot of problems that need fixing, so in many cases it is better to buy them. But in theory, the idea is exactly the same, except that you will just generate what you need.

I think it is important to not look at all these AI tools based on what they can do now, but on what they will be able to do.