jlg1000
Loc: Uruguay / South America
With the new SDXL 9.0, I decided to upgrade Theresa to 3.0 and let her model for me
Wanted to make them look like retro steampunk paintings I used to see on book covers when I was a teenager.
Hope you like the "photoshoot"
AI is fun to play with. The problem I have is that in both MidJourny or Stable Diffusion is getting a full body shot, success rate < 50%. Looks like SDXL 9.0 has the same problem
jlg1000
Loc: Uruguay / South America
Curmudgeon wrote:
AI is fun to play with. The problem I have is that in both MidJourny or Stable Diffusion is getting a full body shot, success rate < 50%. Looks like SDXL 9.0 has the same problem
Well... I can do full body:
I'm putting this imagen on a separarte post also
Not quite. No legs or feet
jlg1000
Loc: Uruguay / South America
Curmudgeon wrote:
Not quite. No legs or feet
Right... trying to do better
jlg1000 wrote:
Right... trying to do better
Seems to be an issue with most, if not all, of the AI software. My best luck has been using: full body shot, standing and changing the orientation to portrait. I sometimes get 2 out of 4 images in MidJourney and not quite that good in Stable Diffusion. I have given up on Firefly
jlg1000
Loc: Uruguay / South America
Curmudgeon wrote:
Seems to be an issue with most, if not all, of the AI software. My best luck has been using: full body shot, standing and changing the orientation to portrait. I sometimes get 2 out of 4 images in MidJourney and not quite that good in Stable Diffusion. I have given up on Firefly
I think I've got it.
You need to give space to the model.
Please explain how to give more space to the model. I use portrait orientation.
jlg1000
Loc: Uruguay / South America
Curmudgeon wrote:
Please explain how to give more space to the model. I use portrait orientation.
By selecting the adequate WxH ratio for the desired composition.
For example:
1) a full body man or woman needs about 1:2 ratio.
2) Because the model I've chosen has 512px resolution, that means that *one side must be 512 px*
3) Therefore, the other side must have 512 x 2 = 1024 px (aprox.)
Therefore, by choosing 512 px W x 1024 px H, I get the full body composition.
There are other techniques, like segmenting, which I'm investigating right now.
How are able to get the same face and different body positions? I can never get the same face with Adobe Fire Fly. I would like to know if you don't mind sharing the information.
How are able to get the same face and different body positions? I can never get the same face with Adobe Fire Fly. I would like to know if you don't mind sharing the information.
jlg1000
Loc: Uruguay / South America
Bill Hancock wrote:
How are able to get the same face and different body positions? I can never get the same face with Adobe Fire Fly. I would like to know if you don't mind sharing the information.
Sure I can.
I'm running open source Comfy UI which in term interfaces with a highly modded Stable Diffusion version in my own Linux computer. I even wrote pieces of the software.
Please note: I do not use any of the "online services"
jlg1000
Loc: Uruguay / South America
If you want to reply, then
register here. Registration is free and your account is created instantly, so you can post right away.