art

Non-Fungible GANscapes is a series of digital landscapes generated using a Generative Adversarial Network and distributed as non-fungible tokens. The works were trained on curated landscape imagery and produced as machine-generated scenes that sit somewhere between photographic memory and synthetic terrain.

The project was originally developed as part of a multimedia art course I taught online at Saint Xavier University. It was conceived as a practical example of what emerging digital art might look like when generation, authorship, and distribution are all mediated by software. Rather than treating AI as a stylistic filter, the system itself was the medium.

Each landscape is the result of an adversarial process: one network proposing images, another rejecting them, iterating toward coherence. The outputs are not composed or edited by hand. Selection happens after generation, not during it. What’s authored is the dataset, the training process, and the constraints under which images are allowed to emerge.

The resulting works were minted and exchanged on OpenSea, situating them within an early NFT ecosystem where questions of originality, ownership, and reproducibility were still unsettled. In that context, the project functioned both as artwork and as a probe: what does uniqueness mean when images are statistically derived rather than explicitly composed?

At the time, AI-generated art was beginning to attract broader attention as a potential direction for digital media, particularly within NFT markets Cointelegraph. Non-Fungible GANscapes was an attempt to engage that moment critically and materially, by putting generative systems directly into circulation rather than speculating about them.

The series treats landscapes not as places, but as probabilities—images shaped by data, taste, and constraint, fixed only at the moment they are observed and exchanged.