Daniel Russellさんのイラストまとめ

Tried to generate some @HamiltonMorris portraits with CLIP ViT-B/16, but it turned out so vague, it almost looks more like me, instead. Time for a new profile pic 🥳

#generativeart #digitalart #aiart #ai #machinelearning #generative #clip

left: some screwed up perlin noise
right: some iterations into something about beksinki rainbow, I don't recall

See the resemblance?

I kinda like em

Mona Lisa init image -> `lisa frank`

#generativeart #digitalart #aiart #ai #machinelearning #generative #gan #clip #lisafrank #monalisa

Thanks to @classpectanon for the text prompt idea! That one usually turns out quite pretty🙂

First image was creating with the typical way of guiding VQGAN with CLIP, towards a text prompt of `a witching hour` (+ others).

The next 3 images were projected into a VQGAN who's parameters had been fine-tuned to recreate the features of the original "style" image

Heeeeere's William!

Training VQGAN params into a `stephen gammell` transformer
Sad Mona (early iteration), then happy Mona (150 iterations later)

Wanted to try something with more details and various colors, compared to the sunny clouds one.
Optimized VQGAN's params toward a prompt related to a painting of a demon in Athens (from @jbusted1)

Still a little in shock of how well this works