AI Art - Examples
Posted: April 14, 2023, 1:57 pm
I didn't want to clutter up the other AI discussion thread so will use this one for various examples.
When it comes to prompts, usually you are less using regular sentences to describe what you want in the picture as opposed to more specifics like
"wearing a hoodie, blue eyes, city background, oil painting" so I saw this one prompt over on CivitAI that was:
with these results: https://civitai.com/images/471229?perio ... tId=139157
I tried the same prompt in two other checkpoints, here's what I got:
model: animatrix_v13
I cut the resolution in half but still good details, the original image would be clear at 4K but they're about 7mb in size.
ok results. she has dirty clothes but really isn't holding out a cup and looks more like a model. No dirt. Definitely doesn't look malnourished.
tried another checkpoint: (revAnimated_v121)
Gets the clothes right, starting to hold out a cup. A little dirty and definitely more crap on the ground, the coins are in the bottom left of the picture.
Still, didn't seem pathetic enough for the description so I used the same checkpoint - revAnimated_v121 and added "snowing, wearing hoodie, baseball cap, snow on the ground" to the description and got this:
Nice, dirty face, looks less like posing for a picture.
If I were to continue work on this picture, I'd go to inpaint and change what looks like a coffee cup to maybe something like a Styrofoam cup with no lid and then fix the feet of that guy walking in the street (bad ankles)
The point of this is that even typing some descriptive regular sentences, you can get a starting point for an image. as for the body, malnourished is hard to genrate, you'd probably have to add (((skinny))) to the prompt, the ((())) adds emphasis in whatever is between them (skinny) less emphasis than ((skinny)) for example.
When it comes to prompts, usually you are less using regular sentences to describe what you want in the picture as opposed to more specifics like
"wearing a hoodie, blue eyes, city background, oil painting" so I saw this one prompt over on CivitAI that was:
The original example used this checkpoint https://civitai.com/models/36732/sb250kplA young girl sits on the side of the street, her dirty clothes barely covering her thin frame. She holds out a small cup with a few coins in it, hoping for the generosity of passersby. Her eyes are downcast, filled with sadness and desperation. She looks around nervously, avoiding eye contact with those who walk by. The city bustles around her, but she remains isolated and alone in her poverty. dirty torn clothes , single cup, malnourished
with these results: https://civitai.com/images/471229?perio ... tId=139157
I tried the same prompt in two other checkpoints, here's what I got:
model: animatrix_v13
I cut the resolution in half but still good details, the original image would be clear at 4K but they're about 7mb in size.
ok results. she has dirty clothes but really isn't holding out a cup and looks more like a model. No dirt. Definitely doesn't look malnourished.
tried another checkpoint: (revAnimated_v121)
Gets the clothes right, starting to hold out a cup. A little dirty and definitely more crap on the ground, the coins are in the bottom left of the picture.
Still, didn't seem pathetic enough for the description so I used the same checkpoint - revAnimated_v121 and added "snowing, wearing hoodie, baseball cap, snow on the ground" to the description and got this:
Nice, dirty face, looks less like posing for a picture.
If I were to continue work on this picture, I'd go to inpaint and change what looks like a coffee cup to maybe something like a Styrofoam cup with no lid and then fix the feet of that guy walking in the street (bad ankles)
The point of this is that even typing some descriptive regular sentences, you can get a starting point for an image. as for the body, malnourished is hard to genrate, you'd probably have to add (((skinny))) to the prompt, the ((())) adds emphasis in whatever is between them (skinny) less emphasis than ((skinny)) for example.