shvrdavid posted at 4:48 PM Thu, 30 November 2023 - #4478399

I have had many things added to Poser, and many things shot down that I suggested. And AI isn't going to happen anytime soon in Poser, or on your PC locally. Running Stable diffusion, etc, is no different than running any other program. Because it is only a program, and there is no AI in it, at all. Zip, nada..... 

How people can not understand that, is beyond me..... Your not running Ai in your pc locally, or your phone, etc. That isn't possible because it doesn't have the processing power or the storage to sort the trillions of findings. Your running a small program, nothing more.

If you have suggestions for filters, things to improve workflow, etc, that's great............. Simply claiming AI will........ Is not...

Poser 13 already has things to clean up images, etc. And it even has things I asked for in there to do so. 

Presentation is everything if you want something added to Poser, and I am telling you that because I have years of experience of getting things added to Poser.

You have to present things in a way programmers understand. You have to know what your talking about. And the Devs have to see a benefit in adding it without breaking the bank. The only thing you have really shown is cleaning up images. And Poser already has that in it. That will undoubtedly see improvements in the future. That also has nothing to do with AI, at all. Tensor cores (AI machine learning cores) are rarely used for filters of any kind, because that isn't what they do....Cuda and CPU cores are superior at filters and complex math. Using the GPU only to do something in a program presents a lot of issues your missing as well. Not everyone using Poser has a Nvidia GPU.... And no one on a Mac does, basically.....

Whatever gets added, has to work in all those cases.....

Saw the RedPhantoms post, so thought I would answer as long as it is productive, which I think ought to be possible, even if people disagree. 

Whether it is true AI or not, it is referred to as generative AI, maybe they will create a better word for it someday.

Also I think the issue is a lot seem to judge it based on what they see now, rather than where things are heading, and also that these things take time to develop, so starting doing it in 10 years makes little sense. 

There are lots of examples of how these AI's are starting to get integrated into programs like PS, Blender etc. And I think we can be sure that if they do it, then we will see it for all the other applications as well, like Unreal, Unity, 3ds max, Maya etc.

I won't link the Blender ones because they are quite technical and rather boring if one is not using an app like that.

But you can see some of the stuff they are working on here: Blender AI

That you say you struggle running AI with a 4090 I don't understand im running it fine on a 1080ti 12gb. But could even run it on my 1660 super with 6gb. Also I think we can be pretty sure that Nvidia etc will develop chips especially for AI at some point in the future, given the popularity of it.

I found this on Reddit, which you could say is extremely relevant for Renderosity:

Hello everyone, I’m Thanos, the Product Marketing Manager of CGTrader, a 3D content marketplace. I wanted to share CGDream.ai with you, a product utilizing SD models that we’ve been working on and recently launched, as I am seeking precious community feedback. What it does is blending 3D models with SD to create 2D visuals, offering users total control of their AI images. As it's still in beta, we're really keen to get your thoughts and feedback.

With CGDream you start with a 3D model from our library that will guide your AI generated visuals. This means you get to control your scene and your image, from the positioning and camera angle to the main subject and concept of the image. Guiding the AI with a 3D model also means that you always get the visuals you imagined, consistency in your image generations, and you can even use your own 3D models (upcoming feature) to generate visuals with them. We also made it super easy to use with no background in 3D needed.

Also it should be noted that 3D models used in CGDream.ai are not used for any kind of training, but merely used to guide the image generation. Currently, our 3D model library is limited, but we're actively working to expand it.

This is the list of features of CGDream.ai:

- 3D to image generation, a new way to generate images with 3D models with just some text prompts

- Numerous filters to choose from to give your images a distinct look and feel

- Our Dream function that uses AI to turn simple prompts into lively and exciting ones for even more interesting results

- Upcoming feature, the ability to upload your own 3D models to generate images

We're aware of some bugs, like login issues and queue display inconsistencies, and we're working to address these. Plus, we're planning a redesign and improving our mobile experience.

CGDream.ai is free to use and also offers a paid plan. I really want to hear from you. What do you think of its approach? How would you use it? Or even stumbled upon a bug or two? Drop your thoughts, questions, or any feedback you have from me. I'll be around to chat, answer questions, and soak up all your insights.

I appreciate all comments, feedback and opinions. Thank you.

------
This is a pure marketplace, yet they seem to have a completely different approach to what AI can be used for. And they don't even have a program like Poser. So again, these tools will pop up everywhere.

Then you have the 3d library that Stability is working on, which is an open source model with 10 million 3d objects in it. 

objaverse-xl

And you can see a video here of it: Stable 3D

As written beneath the video, all the objects and scene were created in a few hours, so one can only imagine this once this technology gets improved and hardware better.