0.Intro00. Prompts and Parameters01. Process02. Style03. Lighting04. Translations05. Tiling06. Animation07. Outpaint / Inpaint08. Graphic 09 Typography10. Word Play11. Type Design12. Abstract Concepts13. Semi Abstract14. Storytelling15. Experimental PortraitsCommunity ExamplesResources“The technology behind them varies, but the most recent AI image generators use what are known as diffusion models. These work by destroying their training data through the addition of Gaussian noise, and then reversing the process to remove noise from the image.”
About 5 billion images are converted into raw data and out of this cloud of noise new, synthesized images are being created using text prompts. So it’s much more than just merging existing images.
The technical process is hard to grasp but one thing that came to my mind is the idea of Star Trek’s beaming using a transporter. People are being dissolved into sub-atomic particles, into information/data, and then put together again in a different location.
BUT that’s not a good analogy. Better is thinking about Cronenberg’s movie called “The Fly”. A scientist invents teleporting and experiments on his own while a fly get into the teleporter with him. This experiment results in a fusion of a man and a fly on a molecular level.
The AI created a huge dense data cloud and out of that cloud or noise it picks or fuses things you ask for with a prompt.Those results are never copies of existing images, rather new creations by synthesizing image data into a digital, visible format again.
Gave it a try to fuse a fly with a man. Not quite like the movie but nevertheless fascinating.
Software progress in the last 6 months. From left to right Midjourney V1, V2, V3 and the current V4:
Found examples complete different manifestations of the same subject – a storm trooper. It’s a good example to get an idea what’s possible with these tools.
Subtile nuances of the impossible
More than just text prompts…
There are other possibilities expanding on the plain text prompt. One example is to add an image to the text prompt. The AI analyses the submitted image and makes it a part of the generated result.
the source image has a big impact of how the prompt will be generated
There may be new terminology necessary to differentiate this new form of image making. Here’s one attempt to define a new word to separate it from AI “art”.
Play with text and image combos using the letters SG. Both, letters and image mixed have been created with AI previously.
