Stable diffusion is an AI tool for generating images from text prompts or other images.
How to get started:
I found this video helpful the basic process is:
- Install pytorch (docker images are available)
- Download the pre-trained model
- Run the scripts
Playing around
Why not start with some Transformers eh?
I was a bit shocked how quick it was to genereate these and how decent they look. I really like the one on the left.
How about dragons?
This didn’t suprise me much, I included artstation in the prompt to get some more detail.
Knights?
These are okay, I love the helmets but I can tell the prompt I gave could have been improved.
How about something cute and fluffy?
I was hoping for a bunch of baby bunnnies…which I got…but oh my! What’s wrong with them?!
More transformers! On the death star!
All-and-all I think these are downright fantastic!
How about something hard! A star wars fight!
A bit confused but not bad!
How about reaaallly fancy sandcastles?
I really like how these turned out!
LEGO on Mars?
Cherry-picking
Here are some of the ones I liked specifically or found interesting.
Summary
Be specific! There’s a great deal the model doesn’t seem to understand. It understands cute though
Some details will be wrong A lot of specifics will slip through the cracks.