Isaac Karth

I'm a technical games researcher with a particular focus on procedural generation. Right now I'm working at Liminal Experiences, Inc., an early-stage startup that's trying to make game development more approachable to non-programmers.

In early 2023, I earned a PhD in Computational Media from the University of California, Santa Cruz. I conducted my PhD work in the Design Reasoning Lab, advised by Adam M. Smith. My dissertation, Recomposing Procgen, documents a bunch of my research from this time period, including the work on WaveFunctionCollapse (WFC) for which I'm most widely known and the work on procgen poetics that I wish more people knew about.

Before UCSC, I earned an MFA in Arts and Technology from the University of Texas at Dallas in 2013 and a BFA in Digital Media from Kendall College of Art and Design in 2006.

Want to get in touch? Email me.

Featured Papers

Conceptual Art Made Real: Why Procedural Content Generation Is Impossible
Isaac Karth and Kate Compton. PCG @ FDG 2023.
Argues that procedural generation cannot replace human artists, because the mere availability of a generator capable of creating artifacts of a given form irrevocably transforms the audience's relationship to those artifacts. Instead, generative art must be understood as a reified form of conceptual art with a distinctive aesthetics of its own.

Constructing a Catbox: Story Volume Poetics in Umineko no Naku Koro ni
Isaac Karth, Nic Junius, and Max Kreminski. ICIDS 2022. Finalist for Best Student Paper.
Discusses the poetics of story volumes and their expression in the formally innovative visual novel series Umineko: When They Cry.

Neurosymbolic Map Generation with VQ-VAE and WFC
Isaac Karth, Batu Aytemiz, Ross Mawhorter, and Adam M. Smith. FDG 2021.
Combines symbolic (WFC) and neural (VQ-VAE) approaches to directly learn how to generate large, mechanically functional game levels from small numbers of example images.

Generating Playable RPG ROMs for the Game Boy
Isaac Karth, Tamara Duplantis, Max Kreminski, Sachita Kashyap, Vajaya Kukutla, Aaron Lo, Anika Mittal, Harvin Park, and Adam M. Smith. FDG 2021. Best Game or Demo.
A demonstration of a novel pipeline for generating complete Game Boy RPGs. We target the internal project format of the GUI-based game creation tool GB Studio as a machine-readable and machine-writable game description language. Example games here. See also our related paper at IEEE CoG 2021.

WaveFunctionCollapse: Content Generation via Constraint Solving and Machine Learning
Isaac Karth and Adam M. Smith. IEEE ToG 2021.
Reframes the WaveFunctionCollapse (WFC) family of procedural generation algorithms in terms of constraint solving and machine learning, illustrating through experiments the role that each component of the WFC pipeline plays in shaping the generated results.

Preliminary Poetics of Procedural Generation in Games
Isaac Karth. ToDiGRA 2019.
Investigates and attempts to characterize the distinctive aesthetic effects of procedural generation on players in digital games.

Addressing the Fundamental Tension of PCGML with Discriminative Learning
Isaac Karth and Adam M. Smith. PCG @ FDG 2019.
Introduces a means of steering WFC-based game level generation via discriminative learning. This enables artists and level designers to quickly and easily improve level generators by providing small numbers of examples of good and bad output patterns.

Links

Here are some other places you can find me:

My frequent collaborators include Max Kreminski (who made this website), Nic Junius (who did not), Adam M. Smith (my PhD advisor), and Kate Compton.