Today we suffered (still suffering) an horrible heatwave (at 21:00 is still 30ºC in the middle of a storm) so I was all way locked in the bedroom with air conditioning.
Today's "Good evening" toot is not a picture but a song. This was created using #ChatGPT from scratch. That is, creating a python program to write midi files and then composing the song using that program.
Today I tried my best to improve my paper so that the reviewers can better understand my methodology (which, tbh, I myself barely understood when I wrote the paper).
Many people think Picasso said "Good Artists Copy; Great Artists Steal". Ironically, this or similar quotes have been said by many artists and there's no evidence to attribute it to Picasso himself (https://quoteinvestigator.com/2013/03/06/artists-steal/).
Anyway, here's Picasso stealing the Mona Lisa. Made with #StableDiffusion, which was arguably trained with stolen images.
Any signal can be represented as a sum of sine waves so this technically works for #genuary15 prompt (also, not really, and waves in the ocean do not break, but whatever).
For "grid inside a grid" I wanted to create one of those trippy infinite recursive zooms. I'm not super satisfied with the result. The seams are painfully visible and the zoom is anything but constant.
I visited London in 2019 and, coming from Argentina, it was my first time on a city older than the colonisation of America. The first thing I noticed is that these medieval cities are definitely not a grid! It's very easy to get lost on the meandering streets and irregular blocks.
I tried to create optical illusions without success before. "Black and white" seemed like a good excuse to retry it with all my learnings. I'm pretty happy with this one.
The trick is to use alternating prompt to create a base image and then iterate img2img alternating between the "vase" and the "silhouette" prompt. Also, I think the Euler A sampler is too unstable for this; DDIM seemed to work better.
Again a #genuary prompt that I don't know how to follow to the letter but that I used as an excuse to try something new. In this case, tiling patterns.
The last prompt is to "ruin" one of the previous images. I noticed that using img2img multiple times with the same seed and the same prompt tends to generate very weird results.
Bonus point for the accidental NB flag to the left of the image.
Lately I don't have a lot of time or energy to do much, so this one's a bit rushed. In any case, today I learned that it's "reindeer" and not "raindeer".
Well, something different. I was looking through my files and remembered this drawing I did in 2011. I took drawing classes with a friend of my mom's and bought a Genius tablet to learn digital arts. Is not awesome, but not super bad either.
Testing a new style and also Tiled Diffusion, which supposedly allows you to tell the model where to put different things (in this case, a treehouse on the right). It's a bit finicky, though.
I think I'm starting to learn how to use this. Instead of trying to prompt everything at one, I first prompted the background, then added a picture of a turtle and inpainted it to blend in style (not perfect, though). Then opened in gimp and deleted every shell part, added a city and again, inpainted to get the style. Then, another batch of img2img to blend the overall image a bit more and finally do the upscaling.
First light using #StableDiffusion XL and colour me Impressed. This is reusing an old prompt for SD1.5, no negative prompt, no iteration and no cherrypicking.
En aquel Imperio, el Arte de la Cartografía logró tal Perfección que el mapa de una sola Provincia ocupaba toda una Ciudad, y el mapa del Imperio, toda una Provincia. Con el tiempo, estos Mapas Desmesurados no satisficieron y los Colegios de Cartógrafos levantaron un Mapa del Imperio, que tenía el tamaño del Imperio y coincidía puntualmente con él...
Add comment