Is there a $ here?
Is there a $ here?
I’m generating an excessive amount of shapes, textures, and colors and need to release some into the wild… Wild as in physical escapes: t-shirts, hats, coffee mugs, wallpapers for your own screens. Soon, exclusive content for members.
imagining more
creating things not yet created
one definition of reality is: that if it is imagined, it is indeed real… at least at some level.
that can be comforting, especially when considering what is out there that has not yet been imagined.
animation creeps in
a very time intensive engineering project, but a nice, interesting effect.
Nature is my “Prompt Engineer”
Here is the original phone (Google Pixel 6) image taken on a recent hike. Nature has already produced an nearly infinite array of textures that can be inspirational to play with in the context of testing, and even training, an AI, or specifically a Large Language Model (which ChatGPT is an example)
Submitting this image to an AI (in this case MidJourney) for a variation image I need to give the AI an indication of what I would like it to do with the image. That is basically called “prompt engineering”
What is a “Prompt Engineer”
I asked ChatGPT v4 to define this….
A "prompt engineer" is presumably someone who specializes in prompt engineering, i.e., someone who designs, tests, and improves prompts to get the best possible performance out of language models.
MidJourney is an example of a massive GPU array tuned and trained to understand common (English only) descriptions that direct it to “interpret” those commands, sometimes paired with an image (like I have done here) to produce something different.
This is the resulting image.
I was looking for something whimsical, as that often produces a fantastical, but slightly believable image.
After about 10 iterations I was satisfied with this image.
my prompt: whimsical --no people hyper-realistic, professional photography, professional color grading, perfect lighting --ar 16:9
*** Nature, as usually the case, offers us opportunities in so many fields, and AI art is certainly one.
Process
? Process
I've been following your posts like this and would like to know what program you're using.
I'm actually using several programs that do most of their computation on large server farms, so I don't need a personal supercomputer to the work.
ChatGPT V4, MidJourney 5.2, Adobe Photoshop Beta.
Basically the process involves feeding the AI an image, or phrase, getting its initial output, and then refining and polishing from there. Everything I've been posting is usually the 5th + iteration of my original prompt.
I'm doing human animation work for training courses that I don't post here. Also, deep research into focused areas of business strategies that save the clients huge amounts of time, and sometimes discovers novel approaches for their work. I'm rambling here, I know, but these evolved AI tools are, IMHO, right up there with cellphones, spreadsheets and microwave popcorn in their ability to increase efficiency.
tricking the Machine
AI provides a new-level, efficient tool to nearly every aspect of computing. As such, it's healthy to test the beast to see just how well it can avoid traps…because that's simply fun…and useful at some other level we'll ignore.
So, let’s play with what is called “AI Generative Fill”, or “outpainting” where the AI is tasked with filling in the missing part, or edges of an existing image.
Assuming the AI has been trained on the classical images, like the Mona Lisa, I had lofty expectations….
I have the original here on the left, and on the right, the version with half missing….
I asked Adobe Photoshop Beta to fill in the left half and got….
Leonardo would be a little disappointed.
now, it can actually do good work, if not perfectly predicting what we demand….
above is an original AI created image (using guidance from text prompts), and BELOW is additional views created by expanding the canvas horizontally and giving the AI a command to “fill in” the blank canvas based on what already exists.
