HUMBABA (2020-)

HUMBABA is a robopoet written in JavaScript and running in a web browser to read in an authored poem-text (the source) from an external file and re-write it - displayed beside that source - producing an echo or failed copy of the source.

HUMBABA is programmed to search for replacement words to those in my authored source text by sound (first with variable degree of similarity, and then aleatoric selection from the possible replacements) and it tries to fit to a constraint I used in writing the source - to fit a line to a given number of syllables. HUMBABA is asked to match the counted syllables, as I hear them, in the source word. If it does not find a replacement that matches in this way - both in being to a degree homophonous and in syllable count - it leaves the source word unchanged.

The source is therefore treated as a grid or scaffold for the echoic variants made by HUMBABA.

As it composes its echo-poem HUMBABA also listens for words that match (in both the source and its echo-poem - the HUMBABA poem) to a given set of critical, diaristic or oblique commentaries located in another external file. If it finds such a match, it displays that word, and from that word can be found (by rollover) the commentary.

HUMBABA is made with p5.js and uses the RiTa library.

HUMBABA sequences have been published in The New River and Poetrishy.

Notes and a commentary on HUMBABA can be viewed here.

An extended essay on writing with a robopoet such as HUMBABA is available on the Lexicon of Generative AI Art: Writing like a loser (2024).

Extant HUMBABA sequences are accessed from the links below (HUMBABA requires minimum window size of 1210 x 755 px).