Skip to main content
c.pfaffenbichler
Community Expert
Community Expert
January 7, 2024
Question

generative fill – weaknesses in certain areas

  • January 7, 2024
  • 1 reply
  • 203 views

Photoshop 25.3.1 on Mac OS 14.1.2

 

I have been trying to automate using Generative Fill via a Script to split image extensions to circumvent the size/resolution issue. 

(I had originally overlooked that the generated images are supposed to be 2000px x 2000px now as opposed to the 1024px x 1024px I had read about a while back but that does not seem to make a meaningful difference.) 

 

The results appear too problematic to be exactly useful in the areas of little contrast (in this case the sky where noticable borders appear). 

 

I have experimented with overlaps and feathering for the Selections but have not come up with any satisfactory combinations yet. 

Do any of you, who may have utilized the feature more extensively, have made similar observations and, if so, found a »solution«? 

 

Thanks for any insights! 

This topic has been closed for replies.

1 reply

Inspiring
January 8, 2024

It's actually not very nice to look at
we are at the beginning of AI, they will certainly solve this problem too.

 

c.pfaffenbichler
Community Expert
Community Expert
January 8, 2024

One would hope so. 

 

But maybe there is a particular reason why such low-detail regions »mislead« this AI to create slight (but ultimately problematic) misses in luminance?