Edit: Ok folks, the LoRA is in the oven. And immediatly the SDXL version glitched, LOL. Good thing the NAI version seems to be ok, I can retry later. The dataset will be available to everyone, of course. Now let's wait...
Inspired from Moritanik idea, I thought to plagiarize elaborate more on it. The rules are the same: I'm going to make a LoRA from the images that you give me (but for IL/NAI, as those are my favourite models: blunt and cheap).
Post an image in response to the the comment that I'll post under this article, It needs to contain a link to an image.
Images must be TOS friendly, and either your work on civit or another external image generator.
Whatever image you post needs some connection to the previous image (this is not serious business: you decide the connection, and it can be whatever - maybe write something funny to explain it?)
no XXX,
it's not my thingok, it's my thing, but let's keep it reasonably clean: X is your ceiling, really.If you've posted an image you have to wait for another person to post before you can post again.
All images should be SDXL friendly resolutions, I'm not going to resize a load of imageswhatever LOL.
This will be a mess but when we gather some images I'll make the LoRA (if I'm not uber busy or hindered by something marriage-shaped). I'm thinking between 100 and 200 generations?
Buzz tips optionals.
Think word association but with images, any images. Never played any Libellud games?