Consider the composition choices in this photo from Jason.
-
RE: https://mas.to/@drumshaman/116083636261429882
Consider the composition choices in this photo from Jason. An image generator wouldn't ever make these choices. Splitting the frame in half horizontally? The difference in contrast between the leaves in the sun and under-story is so extreme.
It's new to me, fresh. And filled with what's missing from generated images which are always pretty and safe, but almost never ... bracing.
-
RE: https://mas.to/@drumshaman/116083636261429882
Consider the composition choices in this photo from Jason. An image generator wouldn't ever make these choices. Splitting the frame in half horizontally? The difference in contrast between the leaves in the sun and under-story is so extreme.
It's new to me, fresh. And filled with what's missing from generated images which are always pretty and safe, but almost never ... bracing.
@futurebird one thing about machine generated images is that the lighting always seems to be consistently boring. Like a Netflix movie, flat and uniform.
-
@futurebird one thing about machine generated images is that the lighting always seems to be consistently boring. Like a Netflix movie, flat and uniform.
The lighting is boring, or it's "textbook" ... at least for photos. Deeply forgettable.
-
The lighting is boring, or it's "textbook" ... at least for photos. Deeply forgettable.
That is mostly because the prompters are lazy and do not ask for interesting effects. Having a virtual camera that can create any kind of image in any kind of style does make you a creative genius with image composition sklls.
It is fairly easy to get an image that shows similar play with light in undergrowth from a model with good prompt adherence such as Bytedance Seedream V4.5 or similar. IF you have the idea to go and ask for it.
I do claim zero originality or artistic value for the attached image, nor do I want to claim it to be in any way comparable to the other image above.
But I wanted to demonstrate how you can ask for lighting, if you care the slightest bit about the subject, and a model can follow through.
-
That is mostly because the prompters are lazy and do not ask for interesting effects. Having a virtual camera that can create any kind of image in any kind of style does make you a creative genius with image composition sklls.
It is fairly easy to get an image that shows similar play with light in undergrowth from a model with good prompt adherence such as Bytedance Seedream V4.5 or similar. IF you have the idea to go and ask for it.
I do claim zero originality or artistic value for the attached image, nor do I want to claim it to be in any way comparable to the other image above.
But I wanted to demonstrate how you can ask for lighting, if you care the slightest bit about the subject, and a model can follow through.
@isotopp @futurebird I am utterly shocked at the idea that someone who asks a machine to generate a set of pixels is both lazy and lacking in creativity.
-
That is mostly because the prompters are lazy and do not ask for interesting effects. Having a virtual camera that can create any kind of image in any kind of style does make you a creative genius with image composition sklls.
It is fairly easy to get an image that shows similar play with light in undergrowth from a model with good prompt adherence such as Bytedance Seedream V4.5 or similar. IF you have the idea to go and ask for it.
I do claim zero originality or artistic value for the attached image, nor do I want to claim it to be in any way comparable to the other image above.
But I wanted to demonstrate how you can ask for lighting, if you care the slightest bit about the subject, and a model can follow through.
@isotopp @futurebird those images don't help the case. The lighting in both looks very wrong. The black and white image especially.
-
@isotopp @futurebird those images don't help the case. The lighting in both looks very wrong. The black and white image especially.
@isotopp @futurebird I think maybe the reason people don't try to do "creative" lighting is because it makes the image that much more uncanny since the model doesn't have any context for the scene, where light would be, and how it would behave. Just spit balling though. I don't really know why it falls on its face so hard
-
R ActivityRelay shared this topic
