Daz Studio scene with Filament render into a photorealistic scene with Stable Diffusion

Hello everyone, I would like to know how I can transform a scene created in Daz Studio with render filament into a photorealistic render using Stable Diffusion. Not having a powerful PC with Nvidia I can only render with filament so I would like to use stable diffusion to achieve a similar quality render.

I would like to get a similar result as shown by this Youtube video. That is, the scene must remain the same, only the graphics I want to improve.

https://www.youtube.com/watch?v=8wAR3Cr0BZI

 

I would really appreciate any advice or experience you could share on how to make the rendering better with this technique.

Thank you very much in advance for your help!

Comments

  • cain-xcain-x Posts: 186

    You may need to take a look through civitai to see if there any models or LORAs that may help towards your goal, then test the hell out of it.

  • WendyLuvsCatzWendyLuvsCatz Posts: 38,168
    edited May 10

    if you use it in ControlNet using Canny or Depth and maybe OpenPose you should get pretty good results

    on single images it is easy to get awesome results, video less so as temporal consistency is not retained even witn the same seed

    for an easy to use SDXL image generator, Foocus can run on Nvidia cards with as little as 4GB of VRAM for images 1024x1024 and other preset aspect ratios maintaining that pixel size and has a Canny image input option 

    https://github.com/lllyasviel/Fooocus

    download the zip that suits your hardware 

    run the bat file, it will download and install everything needed in a self contained folder 

    you can add Loras too

    if you don't have the hardware you can run it in Google Collab 

    Post edited by WendyLuvsCatz on
  • pigna8pigna8 Posts: 1
    edited June 20

    WendyLuvsCatz said:

    if you use it in ControlNet using Canny or Depth and maybe OpenPose you should get pretty good results

    on single images it is easy to get awesome results, video less so as temporal consistency is not retained even witn the same seed

    for an easy to use SDXL image generator, Foocus can run on Nvidia cards with as little as 4GB of VRAM for images 1024x1024 and other preset aspect ratios maintaining that pixel size and has a Canny image input option 

    https://github.com/lllyasviel/Fooocus

    download the zip that suits your hardware 

    run the bat file, it will download and install everything needed in a self contained folder 

    you can add Loras too

    if you don't have the hardware you can run it in Google Collab 

    Using controlnet with canny can you tell me in this case which parameters are recommended regarding cfg, denoising and seed?

     

    Post edited by pigna8 on
Sign In or Register to comment.