July 23, 2026
Enhance? More like invent
What happens when the information runs out
AI gave his brother fake eyebrows — and the comments had zero sympathy
TLDR: The writer discovered that when old photos are missing detail, AI doesn’t recover the truth — it invents a believable version, right down to the wrong eyebrows. Commenters mostly reacted with sarcasm and “that’s literally how this works,” turning a thoughtful warning into a debate about fake certainty.
A sweet, reflective story about restoring old family photos turned into a mini comment-section cage match over what people think AI photo magic actually does. The writer tried using ChatGPT to colorize black-and-white pictures, including one of his brother from the 1970s and another of a road crew from the 1920s. The emotional gut-punch came when the tool invented details that were never there — most memorably, his brother’s eyebrows. Suddenly the photo looked less like memory and more like a stranger wearing it.
That hit a nerve, but the crowd’s reaction was a mix of snark, shrugs, and “well… obviously”. One of the funniest drive-by jokes was basically, did you forget to type "enhance"? Others compared it to those old crime shows where a blurry blob somehow becomes a perfect face after a dramatic zoom. The strongest hot take: this is not a shocking failure, it’s the whole point. Several commenters said if you ask a machine to fill in gaps, it will do exactly what a human restorer would do too — make a plausible guess.
Still, that didn’t stop the low-key drama. Some readers were sympathetic to the larger point that missing information is just gone, no matter how confident the software looks. Others sounded almost baffled that anyone expected anything else from “AI restoration.” In other words: one man saw a haunting lesson about memory, and the internet replied, welcome to guessland.
Key Points
- •The article tests AI colorization on personal and historical black-and-white images to examine what the technology can and cannot recover.
- •A higher-quality rescan of a 1976 negative produced better tonal detail but still did not reveal the brother's eyebrows, indicating the information was not captured in the original image.
- •In a historical road-crew image from east-central Illinois, reference photos helped ChatGPT plausibly color bricks and environmental features.
- •The article reports that ChatGPT invented uncertain details, including facial features and sign lettering, when the source material lacked enough information.
- •The author groups image information into three categories: verifiable surviving details, reasonable inferences, and details that are lost or were never recorded.