Herewith some rather startling art produced by a latent diffusion model implemented in a Colab notebook from @multimodalart. In each case the caption is the phrase that I gave the model to use as the seed for generating the image. I have to remind myself that these images didn’t exist until I generated them today.
I started with a few “X in the style of Y” runs:
Then I tried a few more ambiguous phrases:
I particularly like the final one.
I’ve played around with quite a lot of ML-based stuff over the last couple of years, but the relative ease with which I was able to do this has me kind of shaking my head in disbelief. We’re a shockingly long way from MNIST now.
Via The Checkpoint newsletter