In a previous post I touched on the “portrait debacle”. Basically, we don’t really know what the mathematician Adrien-Marie Legendre looked like, because no authenticated portrait survives; all we have is a caricature and a rough sketch of uncertain provenance.
Well, I think we now know what we looked like, thanks to AI. Here you go:
How did I create this image?
First, I uploaded the following image into the AI Copilot:
And then I uploaded this ghastly picture:
It is the only authenticated portrait of Legendre known.
I then asked the AI to “recreate” what Legendre looked like using the following prompt:
Create a portrait of Adrien-Marie Legendre circa 1810, painted in the manner of a refined French academic portraitist of the early nineteenth century, in the tradition of François Gérard. Half-length formal portrait, three-quarter view, neutral studio background, oil on canvas.
Use the surviving sketch as the primary source for facial structure: broad high forehead, pronounced brow ridge, deep-set eyes, heavy eyebrows, compact nose, full cheeks, slightly rounded jaw, mature face with lines appropriate to a mathematician in his late fifties. Use the caricature only as secondary evidence, preserving a suggestion of the distinctive high, outward-sweeping hair, but rendered realistically rather than exaggerated. Hair grey-white, thinning on top yet still full at the sides, with a characteristic upward flare that hints at the caricature without becoming comic.
Depict an intelligent, thoughtful scholar of the Napoleonic era. Dress him as a respected member of the French scientific establishment: dark tailored coat, high white cravat, waistcoat, understated elegance, no decorations or medals. Hands may rest on a table beside a few sheets of mathematical notes and a bound volume, treated as subtle accessories rather than the focus of the composition.
Style: early nineteenth-century French oil portrait, painterly brushwork, rich but restrained color palette, warm flesh tones, careful modeling of light, dignified and historically authentic. Avoid photorealism, avoid modern digital-art aesthetics, avoid caricature, avoid idealization. The goal is a plausible museum-quality reconstruction of Legendre’s appearance based only on the supplied sketch and caricature, balancing documentary likeness with the conventions of French portrait painting around 1810.
I am sure other people can use other prompts, with the same “seed” portraits, and get other results.
Is this what Legendre actually looked like? We can never know for certain. But if I were shown this portrait in a museum and told it was an authentic painting of Legendre from around 1810, I would find it entirely believable.
