Each word is converted to a number, and that number lands on either a Wikimedia file or a Stack Exchange comment. The result gets pulled apart and rebuilt into a new prompt.
letters become numbers, numbers become files
Assembled Prompt
How the numbers work: each word's own letters (a=1…z=26, weighted by position within the word) produce a number. That number's parity picks the source for that word — a Wikimedia file, or a comment from one of about three dozen non-technical Stack Exchange communities (Cooking, Worldbuilding, History, Philosophy, and so on) — then spreads across that source's ID space to compute a starting point, and scans forward from there until it lands on something usable. Same word, same number, same source, same starting point, same result, every time. Neither API has a "seed" parameter, so this direct ID targeting is the closest real mechanism for consistency; it holds as long as that specific file or comment isn't later deleted on its platform.
Wikimedia descriptions get cleared of cataloguing jargon (collection, digitized, copyright, accession, and so on) and reduced to a real noun phrase, anchored with an article, rather than left as a bare word list. Stack Exchange comments are used closer to as-is, since they're already grammatical human writing — just cleaned of a leading @mention and capped in length. The final line skips colons and dashes on purpose — that punctuation reads as a caption, which is part of what pushes image and video models toward rendering the words as on-screen text instead of a scene.