One of the most prominent generative AI tools that has been making headlines in the last few months. Stable Diffusion is specifically meant to create fully artificial, photo-realistic images. If you’ve ever seen one of those “all of these people don’t exist” montages, it was almost certainly using images generated by this tool.
There’s a ton of models and LoRa’s for it that can create a pretty wide variety of things. I use it for creating on-the-fly watercolor scenes during D&D sessions for my session journal.
These were deities for a homebrew campaign, and the DM had already provided their domain, element, and symbol (i.e. war, fire, stallion). I usually just generate 4 images at a time (only takes a few seconds on a 3090) and pick the one I like the best. Sometimes I’ll generate 2-3 sets of 4, but not often if I don’t have a clear idea of exactly what I’m looking for.
If it’s something really specific I need, I could spend hours using in painting and various noise/models to get what I want.
The previous linked image was much the same process, though the prompts were more detailed as the other players had provided more information on their character’s appearance.
Who are they and what do they do?
One of the most prominent generative AI tools that has been making headlines in the last few months. Stable Diffusion is specifically meant to create fully artificial, photo-realistic images. If you’ve ever seen one of those “all of these people don’t exist” montages, it was almost certainly using images generated by this tool.
There’s a ton of models and LoRa’s for it that can create a pretty wide variety of things. I use it for creating on-the-fly watercolor scenes during D&D sessions for my session journal.
https://pixelfed.social/i/web/post/592814040992006978
Those images are awesome. How many tries and time did you need to get such good results?
These were deities for a homebrew campaign, and the DM had already provided their domain, element, and symbol (i.e. war, fire, stallion). I usually just generate 4 images at a time (only takes a few seconds on a 3090) and pick the one I like the best. Sometimes I’ll generate 2-3 sets of 4, but not often if I don’t have a clear idea of exactly what I’m looking for.
If it’s something really specific I need, I could spend hours using
in paintingand various noise/models to get what I want.Edit: oops, I was thinking of a different montage I did recently: https://pixelfed.social/i/web/post/595611323719481231
The previous linked image was much the same process, though the prompts were more detailed as the other players had provided more information on their character’s appearance.
They ushered “AI” even before ChatGPT. It was “The” rad back in August 2022.