Member-only story
I Tried Generating AI Images in the Style of Thomas Kinkade (And The Results Were Amazing)
Here’s my success story with Graiphics software
Checking my email at Fuddruckers earlier this evening, I noticed a sales email promoting some kind of AI software. I almost deleted it. I get a ton of spam emails, but something made me open and read the offer — and I’m glad I did.
I am now a Graiphics customer.
Graiphics is an image-generating software platform that allows users to spin up incredible images in a matter of seconds.
It’s awesome.
So, how does it work? you ask.
Well, I’m no techie, so I will let Graiphics explain:
Graiphics runs what is known as a Latent Diffusion Model (LDM) which can be very complex, what follows is a simplified explanation of how they operate.
These models are trained by feeding it images along with a description then step by step, these images are turned to random noise (think static on an old television). This training process means the AI network understands how to take an image and then turn it into a completely unrecognizable pattern of noise.
The inverse applies when generating an image. The system generates a random pattern of noise and then step by step this noise is removed to create the final image.
Because the AI understands what things are and their context via the training process, this step by process allows it to iteratively move the noise closer and closer to the desired output.
The preceding text is from their Q&A inside the members’ area. If you’re interested in checking out the software, I’ll provide a link at the end of this article.
For now, let me show you the magic.
There are a lot of options, variables, and requests you can punch into the software to generate images. I tried all sorts of combinations. Some didn’t work out all that well, but some were amazing.
I found the worst luck with people. But objects and landscapes? That’s a different story. Some of my results are incredible.
