In collaboration with Won Pound for his forthcoming album release via minaret records I was recently commissioned to lead an expedition into latent space, encountering intelligences of my own haphazard creation.
A word of warning:
This and subsequent posts as well as the GitHub etc. should be considered toy projects. Development thus far has been results-oriented, with my git HEAD following the confusing and exciting. The goal was to make interesting artistic assets for Won’s release, with as little bandwidth as possible devoted to overthinking the engineering side. This is a fun role-reversal, typically the things that leave my studio look more like brushes than paintings. In publishing this work, the expected outcome is also inverted from my typical desire to share engineering techniques and methods; I hope my sharing the results shifts your perspective on the possible ways to bushwhack through latent space.
So, with that out of the way the following post is a summary of development progress thus far. Here’s a demo:
There are a few repositories associated with this work:
GANce, the tool that creates the output images seen throughout this post.
Pitraiture, the utility to capture portraits for training.
If you’re uninterested in the hardware/software configurations for image capture and GPU work, you should skip to Synthesizing Images.
The first revision of this project was shipped in November of 2020, but the subsequent redesign was commissioned and completed the following summer in 2021. This post primarily a journey through that second revision, and it’s publication comes some time after the deliverable was shipped to the client.
Engineering requirements that arrive downstream from artistic intent are my favorite constraints to work inside of. It forces the engineer to assume the role of the artist, considering the feelings and ideas that will be communicated to the audience with the piece. The engineer also has to become an audience member to understand other factors about how viewing will take place, if the environment will change such that the piece needs to respond in kind. The space in between these to roles needs to be projected into the standard space of product requirements, weights, tolerances, latencies etc. that are common in the profession.
As a part of my freelance practice, interdisciplinary artist Sara Dittrich and I recently collaborated on a series of projects, adding to our shared body of work. The most technically challenging part of these most recent works was a component of her piece called The Tender Interval. I urge you to go read her documentation on this project, there is a great video overview as well.
Two performers sit at a table across from each other, above them is an IV stand with two containers full of water. Embedded in the table are two fingerprint sensors, one for each of the people seated at the table. Performers place their hands on the table, with their index fingers covering the sensors. Each time their heart beats, their container emits a single drop of water, which falls from above them into a glass placed next to them on the table. Once their glass fills, they drink the water. Optionally, virtual viewers on twitch can take the place of the second performer by sending commands on twitch that deposit water droplets into the second glass.
Design and manufacture of table and this insert were completed by Sara Dittrich
The device responsible for creating the water droplets (the dripper) ended up being a very technically demanding object to create. The preeminent cause of this difficulty was the requirement that it operate in complete silence. Since the first showings of this piece were done virtually due to the pandemic, we were able to punt this problem and get the around noisy operating levels of V1 using strategic microphone placement. However, this piece would eventually be shown in a gallery setting, which would require totally silent operation.
The following is a feature overview and demonstration of the completed silent dripper:
If you’re interested in building one of these to add to your own projects, there is a github organization that contains the:
Per usual, please send along photos of rebuilds of this project. Submit PRs if you have improvements, or open issues if your run into problems along the way.
The rest of this post will be a deep dive into earlier iterations of this project, and an closer look at the design details and challenges of the final design. It’s easier to understand why a second iteration was needed after reviewing the shortcomings of version 1, so that’s where we’ll start.
The “forthcoming” project mentioned throughout this post has been released! Check it out here.
Here’s a (long winded) video overview of this project:
Background
Rendered desperate for VRAM by a forthcoming stylegan-related project, I recently had to wade thermistor first into the concernedly hot and strange world of GPUs without video outputs to design a high performance cooler for the NVIDIA Tesla K80.
Too esoteric to game on, and too power hungry to mine cryptocurrencies with, the K80 (allegedly the ‘The World’s Most Popular GPU’) can be had for under $250 USD on ebay, a far cry from it’s imperial MSRP of $5000. By my math, the card is one of the most cost-efficient ways to avail one’s self of video ram by the dozen of gigabytes.
This sounds great on paper, but actually getting one of these configured to do useful work is a kind of a project in, and of itself. I’ll eventually get to this in the aforementioned upcoming post. Today’s topic however, is upstream of all that: the task of keeping these things cool.
