DAZ AI Studio - what it should be

Comments

• richardandtracy Posts: 7,076

  May 31

  And what the announcement says it will be: Blog Post. Regards, Richard.
• no nose Posts: 315

  May 31

  At the very least, I think an interesting use case could have been a built in view-port to image function (similair to that one blender addon)
• marekdopra Posts: 0

  June 1

  Yes. The main advantage of daz3 over AI is that You could detailed arrange any pose without fighting with prompts, and character , scene  consistency. And then there should be AI as helper and enchance renders to photos ( or instead of renders)  
  For now the only thing what I am looking for is to real photo from my scenes - keeping consistency between scenes.
  I made comic nsfw and some AI films based on renders but they are very poor quality compared to ai generated photos and  films from photos.   But i have full controll and only advantage over ai creators is that i  can keep storyline.

  I have tried some AI render enchancements but they change faces and many other things or still are not photoreality,

  if you want examples so find 3dasdman ( but nsfw) 
• FSMCDesigns Posts: 12,842

  June 3

  richardandtracy said:

  And what the announcement says it will be: Blog Post. Regards, Richard.

  It's for the best since the DAZ version is light years behind any other AI apps. I am fully into AI now for image and video generation and haven't opened DS in months and don't see that changing any time soon, especially with how much fun I have having with VEO 3
• richardandtracy Posts: 7,076

  June 3

  Access to AI Studio is gone now. RIP DAZ AI Studio.

  Regards,

  Richard
• WendyLuvsCatz Posts: 39,995

  June 3

  I still think incorporating Wan 2.1 into a plan for animating renders would have been a better option

  I don't care to pay the various Online sites that don't have Paypal and woud rather wait the 4 hours it takes me to do a 5min 720p video

  I would however pay DAZ if reasonable
• FSMCDesigns Posts: 12,842

  June 3

  WendyLuvsCatz said:

  I still think incorporating Wan 2.1 into a plan for animating renders would have been a better option

  I don't care to pay the various Online sites that don't have Paypal and woud rather wait the 4 hours it takes me to do a 5min 720p video

  I would however pay DAZ if reasonable

  After playing with VEO 3, I can't go back to Wan 2.1, LOL
• Snow Posts: 95

  June 9

  I've mentioned this many times before, DAZ should have never taken this route with generative Ai.

  Instead they should have focused on the end stage, the render stage where we can bring our 3d renders to life or add that final touch.

  So instead of hours rendering out the image (in my case) and waiting for a proper DAZ Studio version that supports certain hardware they could have implemented a simple generative ai method to render out in photography quality using a tweaked img to img and upscaler, which would take seconds instead of hours. This would have been revolutionairy since render times and realism always have been a struggle with 3d software. They could even have implemented it without mentioning it's generative Ai that does the work, just a new revolutionary way to render out your 3d scenes. After that they could have looked into animation by also improving realism or adding effects, also at the end stage.

  That way we would stil be the creators and not use generative ai to do it all for us. To me it's either DAZ or Generative Ai prompting, not both unless it's to add a touch of realism or effect after most of the work has been done.

  To me and probably many others here the experience and fullfillment of generating your art with Ai (typing) vs actually creating the 3d scenes in DAZ is completely different

  Anyway, just my 2c...
• sidcarton1587 Posts: 54

  September 3

  So suppose you could have something that watches a set of directories, and when it sees a new render in that directory, it runs a background process that uses an AI upscaler to scale the image (and maybe do other post-processing stuff) in the background, to whatever resolution is desired. Would that be of interest?  Because I think that wouldn't be that hard to do right now, it just would be something that runs outside of DAZ Studio itself. 
• sidcarton1587 Posts: 54

  November 12

  no nose said:

  At the very least, I think an interesting use case could have been a built in view-port to image function (similair to that one blender addon)

   

  Interesting. I was just playing with saving a viewport image. This is an example script from the docs:
  
   

  // If there is a main window
  if( MainWindow ){
  	// Get the active viewport
  	var oViewportMgr = MainWindow.getViewportMgr();
  	var oViewport = oViewportMgr.getActiveViewport().get3DViewport();
   
  	// Capture the contents of the active viewport
  	var oImg = oViewport.captureImage();
   
  	// Get the image manager
  	var oImgMgr = App.getImageMgr();
   
  	// Save the image to a file
  	oImgMgr.saveImage( String( "%1/viewport.png" ).arg( App.getDocumentsPath()) , oImg );
  }

  Now if the viewport image can be saved, it doesn't seem like that much of a stretch to write a short application that watches for new viewport images of this type and runs an AI process, using some variant of image-to-image, to create a fully rendered version of the viewport image. 

  Item #151 in my list of things to automate with AI. :-)

   
• kyoreona Posts: 213

  November 15

  Regarding the combination direction of DAZ and AI, I am eager to focus on the following two aspects
  1. Rapid generation of poses
  Based on the prompt words or the photos, multiple G9/G8 poses can be quickly generated for users to choose from. This will save users a great deal of time. The output of the works will experience an explosive growth, and many people will even attempt to create DAZ comics.
  II. Enhancement of Rendering Speed
  My graphics memory is only 12GB. Can I place 20 characters on the screen and then have the AI directly generate the result based on the viewport, after the AI simulation rendering? In this way, I can save the expensive upgrade cost of the graphics card, and also significantly enhance the richness of my work, rather than just having small stories involving only 2 to 3 people.
• Tiziano Posts: 359

  November 15 edited November 15

  marekdopra said:

  if you want examples so find 3dasdman ( but nsfw) 

  I’ve just found you on DeviantArt. Really great work! This is exactly the kind of art I’d love to create myself as well using DazStudio! 

  Post edited by Tiziano on November 15
• sidcarton1587 Posts: 54

  November 19

  kyoreona said:

  Regarding the combination direction of DAZ and AI, I am eager to focus on the following two aspects
  1. Rapid generation of poses
  Based on the prompt words or the photos, multiple G9/G8 poses can be quickly generated for users to choose from. This will save users a great deal of time. The output of the works will experience an explosive growth, and many people will even attempt to create DAZ comics.
  II. Enhancement of Rendering Speed
  My graphics memory is only 12GB. Can I place 20 characters on the screen and then have the AI directly generate the result based on the viewport, after the AI simulation rendering? In this way, I can save the expensive upgrade cost of the graphics card, and also significantly enhance the richness of my work, rather than just having small stories involving only 2 to 3 people.

   

  For #1, the missing piece is still the underlying AI infrastructure to support it out of the box. Still takes a fair bit of work to set up something like stable-diffusion, especially given the variety of hardware, software, and technical experience people have. Having said that, once it's setup, it is not complicated to script pose generation using a combination of prompt text, sample pose images, and control-net. The tricky part, in my opinion, is the last piece of taking the generated pose file and translating it into something that will be both correct and "pleasing" in DAZ Studio. The latter is more difficult, because even with full control within DAZ Studio, we've all had experiences of native pose files not creating great poses out of the box. They still require a lot of tweaking. 

   

  For #2, the image2image operation, applied against the viewport generated image, seems easy in theory but practically it still is a lot of hit-and-miss. Requires the same level of setup and tweaking as #1 in my opinion. 

   

  For what it's worth, I think you're thinking in the right direction, and AI will eventually get there. :-)

   
• AsuCafe Posts: 161

  November 25

  The key difference between DAZ and other AI users lies in our ability to construct 3D scenes. After setting up a simple scene in DAZ Studio, we enable the AI to read scene data and generate images featuring user-specified body types, distances, poses, and more—all without ever producing six fingers!
• wolf359 Posts: 3,928

  November 25 edited November 25

   

  The key difference between DAZ and other AI users lies in our ability to construct 3D scenes. After setting up a simple scene in DAZ Studio, we enable the AI to read scene data and generate images featuring user-specified body types, distances, poses, 

  Hmmm…Not really an exclusive advantage for Daz studio users
  You can do the same with any 2D or 3D software for creating starter images to guide your AI
  Hence this.

  

  

   

  or this for 2D animation

  

  

   

  

  

  1AXHALO-FORERUNNER-ART1-(58).jpg

  900 x 506 - 317K

  

  

  UR-DIDACTNO-HELMET-.jpg

  900 x 1116 - 789K

  Post edited by wolf359 on November 25
• FirstBastion Posts: 8,047

  November 28

  I'm kinda missing this thread,  and the Daz AI functionality.  Sure it was different from the big players,  but its vivuals did have a decidedly daz render quality whish was just different enough from the sea of sameness on the other big sites. 

   

  Age of the Forerunners,  does look pretty cool.