- This topic has 38 replies, 5 voices, and was last updated 4 months, 3 weeks ago by
.
-
Posts
-
The character isn't defined in the prompt beyond "young woman" so it's to be expected. Interesting that they're so similar with respect to race, hair colour and physique.
Just to be inquisitive I wondered what Google would turn up by using some of your requirements.
Not as close as the generated prompt I used but very close. I would suggest that Flux mines Google images very heavily for its database which would explain the similarity of it's render's ethnicity and overall knowledge of what a "female with sign" constitutes. Standard white female in a standard, non ethnically diverse, non-threatening pose. I guess AI just mirrors our ethnic biases.
Curious about your system specs @tedthelearner. Do you have a "beast"?
Some (Pony) models are trained with them, they are just image quality tags like masterpiece, high quality, etc.
"can take someone elses photo that you like online, drag it into ComfyUI..."
Yeh but not always. With some Civitai images I've tried there's no data with them.
Anyway, If I'm using someone else's image as a base I like to alter it a lot and add my own changes, or what's the point.
AI didnt put blood on the clothing. So atm it cant "think" for itself.
Frankenstein on a skateboard. Thats a random concept. Looks cool though
"I'm sorry girl, but .. those shoes." LOL. Yeh they're sure daggy but it's a 1940/50s style image so they suit. Saddle oxfords would have been even better for her age and that period.
Base on the last few posts Ive figured out what I dislike about AI and what it would take for me to warm up to it.
I saw you changed the shoes without messing with the over all picture. However you changed the girls appearance.
You asked for Frankenstein on a skateboard but did you ask which interpretation of Frankenstein?
My issue is you type in the requires words but it gives you what it thinks you want. Ive seen a few templates that allow you to add multiple pictures but havent tried them.
I could warm up to it if I provided more input and governed the results vs it randomly selecting what I want.
For instance I said a "catgirl in a red dress"
Im sure I could tune this with words but I didnt select the dress, the environment, pose, ect... The computer assumed what I wanted.Can I provide the exact dress, the exact pose, environment, the female model, ect... and have AI kit bash them together without taking massive liberties?
Should a robot need a scope?
- You must be logged in to reply to this topic.
