A New day, a new beginning
Posted: Wed Nov 08, 2023 4:37 pm
Whilst I have found using Perchance to be fun and a source of very good images, I have found it a bit frustrating that tends to do what it wants rather than what you told it to do. So, over the last few days I have been playing with some of the online AI options. Most failed my first major criterium, that they weren't censored, at least not in the adult fetish areas. A lot of the remainder failed the second major criterium, which is that they be free! In the end I was left with the decision I was trying to avoid and accepted that the obvious way forward was to install Stable Diffusion locally. So that is what I did today. It took a while, it used a lot of disk, it needs secondary bits of software (Python, Git, etc), it has some fairly rigid requirements (luckily my home PC met these). By some strange miracle I got through all this and ended up with a clean install at the first attempt, but it doesn't stop there, next you have to download one or more "models", then there's the style.csv to find and install. This it the the point I am at now, I have a usable Stable Diffusion install, some models that I've been and a style.csv that I've just loaded but not played with much. Tomorrow there are other things to try, like Extensions and lots and lots of settings to play with etc.
All that said, it runs, it is usable, I am getting some decent images out of it and it doesn't slaughter my machine, so I am quite pleased at the moment. For the future, well there is still a lot to explore.
So (don't you hate it when people start sentences with "so"), here are four of the images that I generated today. They aren't particularly original or that dramatic, but I thought they were decent as a first effort. Initially my aim was to see if I could get close to one of the photomanipulations I did a long time ago, I picked what I thought was quite a "simple" image to describe for the text:
download/file.php?id=60&mode=view
This was as near as I got to getting the subject to hold her arms out, pretty good otherwise.
I quite liked this as well.
Tried a different text description.
This was generated as a "CGI" type image rather than a photograph (after I loaded the styles.csv file)
So plenty of scope for development and improvement. I'll keep you posted on how things go.
All that said, it runs, it is usable, I am getting some decent images out of it and it doesn't slaughter my machine, so I am quite pleased at the moment. For the future, well there is still a lot to explore.
So (don't you hate it when people start sentences with "so"), here are four of the images that I generated today. They aren't particularly original or that dramatic, but I thought they were decent as a first effort. Initially my aim was to see if I could get close to one of the photomanipulations I did a long time ago, I picked what I thought was quite a "simple" image to describe for the text:
download/file.php?id=60&mode=view
This was as near as I got to getting the subject to hold her arms out, pretty good otherwise.
I quite liked this as well.
Tried a different text description.
This was generated as a "CGI" type image rather than a photograph (after I loaded the styles.csv file)
So plenty of scope for development and improvement. I'll keep you posted on how things go.