A Reddit user recently tasked ChatGPT with restoring and colorizing the world’s oldest photo to test artificial intelligence’s creative edge. Niépce captured the historic masterpiece, known as the 1826 heliograph, nearly 200 years ago.
However, enthusiasm quickly turned to critique as the results veered into historical fantasy, prompting educators and photo historians to challenge AI’s limits in preserving visual authenticity.
ChatGPT was given Niépce’s world-famous image, View from the Window at Le Gras, and asked to colorize and restore it. What followed was a lifelike scene featuring sunlit stone buildings, red-tiled roofs, a church steeple, lush greenery, and a picturesque sky, all imagined with confident flair.
Still, the original image is much more abstract. The iconic triangle-like shape in the center is simply shadow cast on a courtyard, not new architecture. A photography history teacher who has seen the original multiple times described the result as “totally inaccurate,” pointing out that ChatGPT had converted light and shadow into imagined structures and scenery.
Niépce’s original process involved exposing a bitumen-coated plate for days. The result is a grainy, near-monochrome scene with only vague outlines of buildings and a distant tree. That artful ambiguity doesn’t translate to AI.
ChatGPT relies on generative models trained on visual patterns, not on factual preservation. It generated a polished, modern landscape that fits its learned data, rather than conserving the photo’s historical integrity. As one Redditor noted, it was more “AI slop” than restoration.
Current AI models like ChatGPT or image-based tools can only guess what the scene might depict, often inventing details like blue skies or tiled roofs based on modern assumptions.
However, in the future, advanced AI systems trained on historical records, environmental data, and material simulations could potentially generate more contextually accurate restorations. While artistic reinterpretations dominate today, tomorrow’s AI may offer reconstructions grounded in scientific and historical probability.