I’m been working with Code Based Generative Art Systems for 15+ Years. Working directly with a lot of amazing artists. I would also say I've got a pretty good idea of Generative Art and the broad history of the art form.

With the rise of the blockchain, new expressions of code-based generative art have emerged. With that, new terminology.

Long-Form Generative Art

The artist Tyler Hobbs loosely defined Long-Form Generative Art in this post from 2021. I've seen many various definitions. None capture the difference between Long-Form vs. Generative Art that has come before.

Before Long-Form Generative Art - Generative art was generally curated Vs. uncurated. (uncurated might not be in the OED so go with me here).

Curated

Curated means that the artist creates the generative system and tunes it to their liking. The system introduces pseudorandomness and then produces multiple outputs. The artist evaluates each output and picks one or a few from the many. At the dawn of Generative Art, plotters were the most efficient way of capturing the output of a generative art system, hence why many of the early works were curated and captured as plots (and later as fabricated objects).

Uncurated

The flipside of the curated coin being uncurated. The artist creates a system and tunes it so they are happy with any output without any additional curation. WYSIWYG. This was fairly unusual for static outputs of generative art such as plots, prints or images.

However, this form of creating generative art was absolutely common for real-time visuals for performances, concerts and reactive installations. For a music performance, the system was fed the music as an input that was then transformed into real-time synced visuals that were also projected onto the stage.

So why don't we stick with the curated vs uncurated taxonomy? I think there are deeper underlying criteria that we could observe that clarifies the position of long-form generative art.

To define, Long-Form Generative Art into criteria. Is this possible? I've tried to place the criteria into HARD or SOFT. Hard being hard absolutes and soft being, well a bit harder and fuzzier to agree with.

Hard Criteria

LONG-FORM GENERATIVE ART = CODE BASED + UNCURATED + UNIQUELY SEEDED + RANDOMLY SEED + CONSISTENTLY SEEDED + UNIQUE OUTPUTS + DETERMINISTIC +  COLLECTION

CODE BASED

Generative art comes in a lot of flavours. It can use generative systems powered by the wind or the sun, thus avoiding the use of software and computers. For long-form generative art, let's just all agree that the art must be made by a code-based system.

UNCURATED

This seems to be the most obvious hard rule of what makes up long-form generative art - it is NOT curated by the artist. The artist creates the system and does not curate the outputs.

UNIQUELY SEEDED

For LFGA, the art is created with a unique seed. The same seed cannot be used more than once.

RANDOMLY SEEDED

The seed that creates the LFGA MUST be randomly sourced. This could be a transaction hash from the blockchain or it could be some other unique random number / hash. Some previous generative artworks before the blockchain used UTC date/time down to the millisecond as their seed - Does this count as randomly seeded?

CONSISTENTLY SEEDED

The seed that creates the art must be consistently tied to the art. Without the blockchain, this could have been embedded in the metadata of the art, the filename or as a domain name. Coupled with the blockchain, this metadata is tied to the art generally as the transaction hash and is the basis for proving ownership of LFGA on the blockchain.

UNIQUE OUTPUTS

Even with unique seeds, the system may not create unique outputs. For LFGA, the outputs of the system must all be unique.

DETERMINISTIC

When the code to create the LFGA is executed, the code MUST behave the same every single time. You reload the art with the seed and it will look and animate the same. Even with time-based systems, the art must repeatedly progress in the same way over time, EVERY time.

COLLECTION

It’s impossible to state what is the minimum threshold for a collection size to be counted as LFGA. I’ve seen collections with less than 100 and it’s also impossible to predict if a collection will sell. Your intent as an artist could be 500, but you might only end up with 50. To be classified as long-form generative art, I believe a collection needs to have more than one item.

Soft Criteria

OUTPUT QUALITY

A benchmark for LFGA seems to be a consistently high quality of output. This is obviously impossible to define and seems to be more of a goal than an absolute.

COLLECTION QUALITY

Another benchmark for long-form seems to be a consistently high collection. Again this is impossible to define and seems to be more of a goal than an absolute.

But what about “blockchain” in the criteria? I'm on the fence about including this for now.

Also I've heard "short-form" bandied about. This seems to mean "curated" therefore I suggest we rally around "curated" as the term.

Proto-Long-Form

I believe there are prior examples of proto-long-form before the blockchain existed.

Invader Fractal (2003)

From 2003 by Jared Tarbell. Invaders are friendly expressions of numerical magnitude. Using only 15 bits, a mirror, and a little patience, we can render 32,768 unique instances of them. They were sold as unique prints. http://www.complexification.net/gallery/machines/invaderfractal/

Lovebytes (2007)

20,000 amoeba-like characters were generated through a software system and then captured as postcards. This project was a collaboration between Universal Everything & Karsten Schmidt AKA toxi. Karsten confirmed to me that these creatures were unique and uncurated. https://www.universaleverything.com/collaborations/lovebytes

https://www.flickr.com/photos/toxi/albums/72157600122947693/

Faber Finds Generative Book Covers (2008)

A project by Karsten Schmidt AKA toxi from 2008 to create print-on-demand generative book covers. 'the task given was to build a “design machine” which would be flexible enough to generate a very large (theoretically infinite) number of unique designs, one for every single book ever printed in this range, within the agreed boundaries' https://web.archive.org/web/20080901194712/http://postspectacular.com/process/20080711_faberfindslaunch

V+A FOREVER 2009

Another collaboration between Universal Everything and Karsten. A monolith created a never-ending visual through a generative system. Over time, the visuals were captured as 20,000 uncurated postcards.

Written Images (2011)

The Written Images Project from 2011 shows another print-on-demand system. Artists created code systems which made an image which was then dynamically stitched into a book - http://writtenimages.net/press_pack2011_3.pdf

I hope people find these criteria interesting. Perhaps sparking further debate and definition. My only favour is that people do not use the term "short-form generative art" as I feel like "curated generative art" is probably what they mean.

Find me on Twitter for further discussion.