Medieval literature often survives today as text alone—studied, excerpted, and summarized, but rarely experienced as something lived. This project began with a simple question: could modern generative AI tools help translate a medieval literary cycle into a contemporary form without flattening its moral weight or historical structure?
The source material was Reynard the Fox (Reineke Fuchs) by Johann Wolfgang von Goethe, using the 1855 English translation by Thomas James Arnold. The goal was not modernization or parody, but adaptation: to present two chapters of the Reynard cycle as music videos that remain faithful to their logic, tone, and consequence.
The result is a paired set of works:
Reynard the Fox — On the Road to King’s Justice (A Medieval Bardcore Confessional Ballad)
Adapted from Chapter the Eighth
Reynard the Fox — Before King’s Justice (A Medieval Bardcore Trial Ballad)
Adapted from Chapter the Fourth
Both were created using large language models (via ChatGPT v5), image generation (DALL·E 3), and music generation (Suno)—but under deliberate, often restrictive constraints.
Constraint as the Primary Design Tool
Rather than asking what AI could add, the project began by deciding what it would not do.
Each video uses:
-
a single, unmoving illuminated-manuscript-style image
-
no animation, no camera movement, no spectacle
-
subtitles treated as compositional elements, not post-hoc accessibility
-
strict fidelity to the structure and logic of the source chapters
In the trial chapter, for example, it would have been easy to invent drama—trial by combat, a last-minute escape, a heroic reversal. These options were discussed and rejected. The court scene in Reynard the Fox is cold, procedural, and final. The adaptation preserves that.
The visual format reinforces this decision: the image behaves like a legal record, not a dramatic tableau. Nothing moves because nothing can change.
These constraints were not imposed by tooling limits; they were creative decisions actively enforced with the tools.
Simulated Agentic Workflow
To manage those decisions, the process used a simulated agentic workflow: not parallel AIs, but distinct functional perspectives invoked as needed during iteration.
Examples included:
-
a Narrative & Themes perspective repeatedly testing deviations against canon (“Does this alter the chapter’s logic?”)
-
a Manuscript Style guardian enforcing period fidelity and restraint in image-generation prompts
-
a Voice differentiation architect solving a specific music-generation problem: two characters, one voice model, requiring perceptual separation through delivery constraints rather than speaker labels
-
a Subtitle readability role treating text placement as part of the visual artifact, preserving the manuscript’s lower margin as writable vellum rather than cleared screen space
The language model functioned less as a generator and more as a structured reasoning surface—a place to argue decisions through, test alternatives, and deliberately not choose the most obvious option.
Two Chapters, Two Different Problems
Although the same tools and workflow were used, the two songs posed very different challenges.
On the Road to King’s Justice (Chapter the Eighth) is confessional, rhetorical, and unstable. Reynard speaks freely, indicts others, justifies himself, and never truly repents. The musical problem was how to let language run without resolution.
Before King’s Justice (Chapter the Fourth) is the opposite. Speech narrows. Authority speaks slowly. Witness replaces wit. The verdict arrives not as drama, but as record and sentence.
Using the same generative tools under different constraints made the contrast legible. The tools did not produce sameness; the structure did the work.
Reflections on AI and Restraint
This project suggests a different role for generative AI in creative work: not as an engine of excess, but as a collaborator in refusal.
By foregrounding constraint—visual stillness, textual fidelity, procedural pacing—the tools became useful for thinking rather than embellishing. The system improved not through added capability, but through the enforcement of standards it could not bypass. The most important contributions were not outputs, but conversations: about canon, consequence, and when not to invent.
Old texts do not need modernization to feel alive. They need forms that respect how they already work.
Do you agree?
Sources and Attribution
Both works are adapted from Reynard the Fox (Reineke Fuchs) by Johann Wolfgang von Goethe, using the 1855 English translation by Thomas James Arnold.
The original work is in the public domain; these adaptations reshape its language and structure into medieval bardcore ballads without altering their moral force.
Public domain source courtesy of the National Sporting Library & Museum (Middleburg, VA), via the Internet Archive:
https://archive.org/details/reynard-the-fox-1855