Paradise Lost, Book 1
These extracts were used as positive prompts
- “So spake th’ Apostate Angel, though in pain, / Vaunting aloud, but rackt with deep despare: / And him thus answer’d soon his bold Compeer.” [125-7]
- He spake: and to confirm his words, out-flew / Millions of flaming swords drawn from the thighs …” [663-5]
- “There stood a Hill not far whose griesly top / Belch’d fire and rowling smoak; the rest entire / Shon with a glossie scurff, undoubted sign / That in his womb was hid metallic Ore, / The work of Sulphur.” [670-4]
negative prompt
monster, fantasy, sci-fi, extra people, poorly drawn, extra limb, floating limbs, airbrush, photoshop, plastic doll, cropped, out of frame, worst quality, low quality, jpeg artefacts, duplicate, extra fingers, poorly drawn face, extra limbs, cloned face, extra arms, too many fingers, caricature, facebook, youtube, food, lowres, text, error, poorly drawn hands, blurry, extra legs, username, watermark, signature
base model: Stable Diffusion 3.5 Medium
image & prompt_1
image & prompt_2
image & prompt_3
summary
These images are from Ilia’s test using Paradise Lost extracts, his negative prompt, and a set of 20 unlabelled photos to see if AI could interpret the text alone. The base model was Stable Diffusion 3.5 Medium. All training was performed by interfacing directly with the model’s code. The resulting images were a product of a severely overtrained model — especially evident in the second image.
The test outcome shows that some poses resembled the library images, and this was because there were few outputs from multiple training steps. To increase compensation for a smaller dataset some light image modifications were inserted into the training pipeline. One of them was a random rotation of the image, from -15 to 15 degrees. The black borders on the second image is a product of having these augmentations and overtraining the LoRA model. Prompts 1 and 2 produced stereotypical, overly polished religious imagery — an aesthetic I aimed to avoid.
This test was part of the exploration to show AI’s representational limits and its bias toward oversimplification, grotesqueness, or artificial perfection.
Karren Visser 