AI image generation comes for Quarter to Three

I released the dragon and pointed @arrendek towards VQGN+CLIP deep learning image generation. He has been experimenting on qt3 user names. It’s like Frankenstein, but with 2D art.

If you want to try this yourself, here’s what you need to do:

  1. Install the NVidia CUDA toolkit: CUDA Toolkit 11.5 Downloads | NVIDIA Developer
  2. Install Anaconda Python: Anaconda | Individual Edition
  3. Install Git:
  4. Follow these instructions: GitHub - nerdyrodent/VQGAN-CLIP: Just playing with getting VQGAN+CLIP running locally, rather than having to use colab.
  5. This requires a lot of VRAM. With 10Gb, you’ll max out at 448x448 images. To mitigate that, you can use ESRGAN, to use deep learning to upscale the images: GitHub - xinntao/Real-ESRGAN: Real-ESRGAN aims at developing Practical Algorithms for General Image Restoration.

Examples (note it uses a random seed and is different with each running):

python -p "An impressionist painting of a burning cat eating apples on a space station | trending on Artstation" -s 384 384 -o "cat1.png"


python -p "Tom playing an extreme sports game and streaming for an online audience|character design on Behance HD" -s 448 448 -o "tc2.png"


Why “trending on Artstation” and “character design on Behance HD”? Best to check out these articles:


Dave examining a packet of Oreos for @Dave_Perkins


Love it, thanks @Clay

You can do less insane things, should you choose. Here is “A charcoal sketch of me standing on a lawn throwing a ball to a black dog.”


Any prompt requests? I should add that the results are more interesting when there are a lot of adjectives and sometimes longer strings/paragraphs work well.

Dingus standing up in a movie theater, pumping his fist and shouting, “NICE!”

Can this be done with pip instead of Anaconda? Not that Anaconda doesn’t look cool…

Sure, but you’ll need to use venv and create different environments for the image generation and upscaling because they use different versions of python.

“John Many Jars fills his jars with a strange white liquid | Trending on Deviantart”

“A Rembrandt painting of John Many Jars”

“Grandma got run over by a reindeer By Thomas Kinkade”

“Randy worries about the coming of a dark lord from beyond”

Per @Chappers, “Oil painting of people running away when Mellified and Randy started the beard wars”


Mine just looks like a nightmare. Did you add anything to affect the style?

python -p "Dingus standing up in a movie theater pumping his fist and shouting NICE | trending on Artstation" -s 448 448 -o "dingus.png"

“Trending on Artstation” is my go-to for getting better results. I need to explore the other models and modifiers. In general, images look far more chaotic if you don’t use a modifier.

I love how the dark lord one looks like Anderson Cooper meets Homer Simpson.

Screw you all, I’m flipping these as NFT’s and going to live on an island somewhere! :D

Nice! Even reminds me of the aesthetic of the street level scenes in Into Darkness.

python -p "The dark lord is Anderson Cooper crossed with Homer Simpson|trending on Artstation" -s 448 448 -o "dark2.png"


That is wonderful.

“McRib sandwich in the style of Piet Mondrian”



Thinking about lunch @Clay ?

Ok, last one for now…

“Putin dressed up as the Easter Bunny.”


How about one for “Armando shitting on a gravestone in the style of Jacques-Louis David.”

I would absolutely frame that and hang it on my wall.


Same, but in the style of Norman Rockwell.


Looks like it’s drawing from the same source images for the gravestone.

Same, with Norman Rockwell, but added modifier trending on Deviantart: