>>> GAN ⫷glitch⫸ Ting Xu | 幹亭序 <<<

• Generative glitch collage: More than simply a + b + c •


At the apex of Chinese calligraphy, Lantingji Xu, the preface, was written by the most renowned calligrapher in China, Wang Xizhi (王羲之).

During the Spring Purification Ceremony, forty-two literati gathered along the banks of a coursing stream and engaged in a "winding stream" drinking contest: cups of wine were floated on the water upstream, and whenever a cup stopped in front of a guest, they had to compose a poem or otherwise drink the wine. At the end of the day, twenty-six literati composed thirty-seven poems in total and the Lantingji Xu, as a preface to the collection, was produced by Wang on the spot.

In the GTX collection, we create new intricate and expressive relationships between ink, paper, and pixels using Jimi's raw calligraphy work, their GAN offsprings, and Thomas's CHX-like glitch processing, generating unique calligraphy on mint and toasting at the same time.


How it works:

Combining sets of raw layers that range from paintings, calligraphy, and GAN seeds, our algorithm applies an array of image processing magic, creating intricate novel outputs.

We have 102 raw layers separated into four groups. Each final output is formed by mixing and processing four randomly selected layers.

Group 1 contains textures, and it is composed of Jimi's paintings, GAN seeds from mixed custom models, and CLIP Guided Diffusions.

Group 2 contains Jimi's calligraphy work.

Group 3 contains GAN calligraphy seeds from custom models we trained on Jimi's body of work.

Group 4 contains digital grain textures.
Our algorithm starts by selecting one layer from each group, assigning them sizes, opacities, orientations, and other parameters. Then it sets a dithering combination to be applied. We have eight combos formed by combining seven different custom dithering algorithms (3 at a time).

After each dither gets applied to its corresponding layer, images get mixed using different blend modes to help combine them in fun ways. Afterwards, a couple more filters get applied to glue everything together. When all is said and done, final artworks get rendered onto the canvas.

Sit back, open the live view, and enjoy.


Appreciate the art, not the speculation. Pay what you can.

With love, sweat, and tears –drops–
Jimi and Thomas.


Jimi's notes:

I provided the calligraphy representing the extreme spectrum of my expression in brushwork and poetry. I also evaluated the GAN calligraphy outputs, helping Thomas hone the models. During the training process I was horrified and in awe by both the stupidity and sometimes brilliance of the GANs.

Through the layering techniques developed by Thomas, I have also gained new insights into calligraphy, taking it beyond brush, ink and paper. The spectrum of expression in calligraphy gets freed from traditional constraints in GTX. Fx(hash) calligraphy has become brush, code and screen.

Thomas does not read Chinese, but I was happy to let him crop my scrolls into square layers in ways he saw fit with the project's aesthetic. To my surprise, I did not only agree with his visual choices but also noticed how the illiterate crops had many Freudian slips. That is, my poetry made sense fractally. I think this is no coincidence for calligraphers that write their poetry. Calligraphy is more than the sum of beautiful brushstrokes and also greater than the sum of pretty characters. Calligraphy is the composition of space and semantics. Only in turn does meaning takes its place in parts of calligraphy.


Controls:
[s] save as PNG.


Paint / Ink / Photography –> GANs / Diffusions –> p5.js –> Blockchain


Copyright (©) 2022 Thomas Noya and Jimi Wen
Licensed under CC BY-NC-SA 4.0

◉_◉ ◉_◉