Show HN: I thought merging two photos with AI would be a weekend project. Nope At a friend's gathering, someone mentioned wanting to add her late father into a family photo. I figured this would be trivial — modern image models are powerful, just send two photos and ask the AI to merge them. She said she'd tried, but the results were weird and unstable, and she didn't know how to describe what she wanted. I went home and tried it myself. With a well-written prompt and two good photos, it works. But real-world use cases aren't two good photos — it's a modern family photo plus a damaged old portrait, or two old photos from different decades. That's when things fall apart. I looked at existing tools. Most showcase merges between clean, well-lit, modern photos. Nobody was solving the hard version: mismatched eras, damaged sources, different poses, different formality levels. I thought it would be a weekend project — one system prompt and done. After 200+ test iterations I realized stable results require much more than prompt engineering. The main challenges: The AI subtly changes faces during merging. The result looks "similar" but isn't the same person. For someone trying to add a deceased loved one, that's a complete failure. Posture and scale need to match. If the group is sitting on grass and the added person is standing like a giant, it's obviously fake. Casual accessories from a reference photo break formal scenes — sunglasses on the head at a wedding, sportswear at a ceremony. Old photos have low resolution, damage, and no color. Merging, restoring, and colorizing at the same time makes everything harder. It supports headshots, half-body, and full-body reference photos. Old and damaged photos work too — with optional colorization and restoration. Still an MVP. Free to use once per day, no login required. https://ift.tt/1zMxWvi Would love to hear about edge cases that break it. https://ift.tt/1zMxWvi April 2, 2026 at 11:32PM
Tags
Hacker News