Pinned Post
Hey! This is a sideblog for @cadmusfly for presenting images synthesised with various neural network AI image generators. It began with a program called Disco Diffusion, though has since branched out to others such as Midjourney, Dall E and Stable Diffusion.
Disco Diffusion AI Art Prompt: “a beautiful painting of a pirate ship on a vivid sunlit ocean by Ivan Aivazovsky, J. M. W. Turner and Lisa Frank, naval, regency, trending on artstation, gay color scheme”
Neural Network AI image synthesis is a method of programmatically generating novel images based on text prompts such as the above. Because of OpenAI's research of the CLIP neural networks and techniques, Dall E opening up to the public via the waitlist and open source work on Disco Diffusion and Stable Diffusion, neural network image generation has become more accessible to the general public as the technology improves.
At the same time, this does pose interesting and tricky questions as to the nature of art, the role of the user in relation to neural network generated images, capitalism's response to emerging technologies that cause concern to existing industries, the biases of the trained datasets and many more.
I like to think of myself as a director of the neural network AI and a curator of the images I choose to present, but your mileage may vary! While most images are generated solely with text prompts, sometimes I do use my own drawings and art as input and canvas for the generator to edit, and sometimes I edit the generated images afterwards.
This blog primarily started with the use of Disco Diffusion (tag here), which is a free open source technique hosted on a "notebook" in Google Colaboratory Notebooks. Images usually take about 15 to 30 minutes to generate, based on the resolution and other factors.
This is an indepth guide on how to use Disco Diffusion
While this is an extremely basic guide that I wrote on how to use Disco Diffusion for non-tech savvy friends
Other image synthesizers include:
Midjourney, Discord based and geared towards artistic images with an interesting gritty texture
Dall E, the famous and most funded one, website based and best at coherent, well structured images
Stable Diffusion, open source and almost as powerful as Dall E
Neural network image synthesis often struggles with prompts of humans and animals, with Disco Diffusion prone to surreal body horror. The other generators are better, but in any case my generations of people are tagged with #dreams of people.
Due to my interest in Age of Sail media and the history of the Royal Navy during the Napoleonic Wars, I mostly generate pictures of sailing ships and resplendent rainbow gay boats, but I'm quite interested in experimenting and playing around with what I can find!
I'm not an expert by any means on neural networks or neural network image generation, but do feel free to ask questions or submit ideas, and I hope you enjoy looking at these interesting pictures!
Tag list
Informative thread by ai_curio on how AI Art is generated
Analysis of the Neural Network AI Generators I Have Used
Thoughts on how neural network AI image synthesis works and some thoughts on ethics
A link to Are We Art Yet? a discord and collective for ethical AI art creation
The Captain's Dream, an AI Art Comic illustrating the words of 1800s naval novelist Captain Marryat
If you wanna request something, read this first!
















