Playing arround with Sketchup diffusion
-
I am testing Sketchup diffusion.
I took an old model and let it go through the new app sketchup diffusion. the result is quite good but the quality looks a bit like a sketch. I took the Picture to stable diffusion and developped it further with image2image and realistic vision. The result is really great...
-
here a test with interior design:
and an additional pic with twinmotion
-
and a tramwaymodel:
-
That is way cool, can you get multiple views of the same image? or will the AI create a different reult for different views?
-
Very cool Jo-ke, if you used/use prompts, would you mind posting the prompt? Would be very helpfull
-
@solo said:
That is way cool, can you get multiple views of the same image? or will the AI create a different reult for different views?
that's the typical AI problem, always different results.
I hope there will be more control in the future.
Stable diffusion has controlnet, a good tool to create consistant pictures but I have no idea to use it together with sketchup.
Maybe in the future there will be more development
-
@l i am said:
Very cool Jo-ke, if you used/use prompts, would you mind posting the prompt? Would be very helpfull
I've used very simple prompts like: ultrasharp photography of a modern dining room in beige and blue colors, tropical garden outside"
-
still testing the options....
I just started to combine render and AI.
rendering cloth in wardrobes has really bad results, because of low resolution models. If I would take better models for cloth the filesize would rise enormously.
AI does really a goof job in makeing realisic cloth and a bad job in doing the rest in desired materials and colors.
so I took my render with the right setting for my wardrobe and took the cloth of the stablediffusion output and merge them in Photoshop:
unfortunately I can't export my POV Camera to twinmotion. this would be helpfull for later work in photoshop
-
Wow the first one is so impressive
Scarily so 
-
-
another one. diffusion is the loser in this case.
this time twinmotion wins.
all renders optimised in photoshop
-
Why not try the Free Vizcom online ?
So easy to use!
And the first on this concept domain...
Team of 2 persons in free!
No limit of number images
Only limit is one image proposition by render (Pro 4)
From your image B&W
But you can import a 3D model... (obj format)
-
you can of course all detail...or modify by prompt...
InPainting (G)
-
hmmm. I think su diffusion does a better job, what I can see.
-
from your 2D image &W
-
Seems there is no change between original and render ...
-
This one looks better.
My general problem with AI is, that I have to less control over the result.
In my case, I want to sell a shelf or a wardrobe and it has to be a consistant look.
But AI does a wonderfull job in textiles and Lightning.
So I am trying to combine them in Photoshop.
-
Yep even some ai accept 3D input files i am quasi sure this 3D is translated in 2D before any process... in some months maybe this will be changed...
-
mixing AI and render in photoshop is a really good workflow:
especially the textiles come out much better
-
Have you try Krea ai or Magnific ai for enhance the first ai result ?
(Magnific ai has not try version)
Advertisement