How to find every video of your grandparents
There's a frame somewhere on your drive of your grandmother at forty, caught mid-laugh across a crowded kitchen — fifty pixels of a face you'd know anywhere, on a VHS tape that's been copied one too many times. You've meant to find more of her. But more of her is buried in a shelf of four-hour files named Tape_04.mp4, behind an empty room, a birthday you weren't looking for, a Christmas you'd forgotten. Digitizing gave you the footage. It didn't give you a way back to her.
This is the way back: how to find every video of your grandparents by their face, in about the time it takes to name them once. You find your grandmother on screen a single time — any tape, any age — point at her, and the search fans out across every tape you own, gathering her into one place. On the real archive Rediscover was built on, the founder's grandfather turns up in around ninety different tapes. Nobody scrubbed ninety tapes to learn that. They pointed at his face once.
It starts with one frame
Here's the whole move, and it's smaller than you'd think. You don't describe your grandmother, or dig up a sharp reference photo and hope. You find her on screen once — at a christening, at the lake house, across a crowded room — put the cursor on her face, and press a key to name her. See her, point, name. Playback doesn't even stop.
From that single confirmation the archive goes looking for that same face everywhere else: at other ages, half-turned, half-lit, through different cameras and formats, across every tape on the shelf. What it finds, it brings back to one place. And it compounds — every appearance you confirm joins her reference, so the next pass catches her where the last one missed. You do a little naming; the archive does the searching. That's what it means to search old videos for a person instead of scrubbing for one.
One page per person, not a wall of timecodes
What comes back is the part that moves people. Not a raw dump of every frame she's in — a page that belongs to her. One best clip per tape, laid out like a shelf you can sit down and watch: your grandmother at twenty-six, at forty, at seventy, gathered from tapes that were never near each other in the box. A relative scattered across dozens of tapes and thousands of clips becomes something you can hold in a single evening.
That is what turns a hundred anonymous files into a person. On the archive Rediscover was built on — 105 tapes, 134 hours, 13,054 faces found — the founder's grandfather has a page that runs across roughly ninety of those tapes: one man on footage reaching back to the 1980s, on a shelf you can page through. Ask to see Grandpa and you get Grandpa, not the question of which tape Grandpa was on.
The naming has to be human
There is a do-it-yourself version of this, and it's worth being honest about. Free, self-hosted tools will scan a library and cluster faces on your own machine, if you're comfortable running a server and don't mind the upkeep. But they do the one thing that quietly breaks an old-tape archive: they guess who people are. On a fifty-pixel VHS face, an automatic guess is just a guess wearing a confidence score.
And a wrong guess doesn't just mislabel a clip. Because the search learns each person from the faces tied to their name, one bad tag teaches it to confuse two people from then on — and the highest risk is exactly where families feel it most: a grandmother and her sister, two brothers who sit a hair apart in the math. So Rediscover draws the line there. Software finds where a face is; a person who knows the family says who it is — you look, you name, and because a human said it, it holds. This isn't theory. An early experiment in automatic face-matching failed outright — there's a $24 receipt to prove it — which is why every identity here is confirmed by a person. Restoration can sharpen a grainy face, but it never invents one, and the archive never invents a name.
When she was always behind the camera
Some grandparents are hard to find by face for a tender reason: they were the one holding the camera. The grandmother who filmed every birthday barely appears in any of them. Face search alone stalls on her — so you find her another way, through the people she was always beside.
Naming faces draws a second thing almost as a side effect: a map of who your family was to each other. Once people have names, the archive knows who shared the screen with whom, and sharing a tape usually means sharing an occasion — a holiday, a trip, a Sunday. Pick your grandfather and you see who he was most often filmed next to, ranked, the thread running on to the next person and the next. Follow it, and the camera-shy grandmother surfaces in the rooms she was always in, even when the lens rarely turned her way.
Start with the tapes you already have
You don't re-digitize anything. Bring the files a service already handed back — Legacybox, iMemories, Costco, the local shop, a drive in a drawer — or the original tapes, and the transfer is handled: VHS, VHS-C, Hi8, MiniDV, Betamax. Everything is catalogued with a real frame pulled from inside each file, so your grandmother's page opens on her face, not one of a hundred identical blue rectangles.
Then it's built to keep. Every original master is copied into a sealed cloud that's yours alone — hashed and verified, never mixed with another family's, never used to train anything — for about a dollar a month for a whole hundred-tape family. That's the copy that outlives the drive in the closet and the tape it came from. Your originals stay yours, and you can walk away with every file, any time. The point was never storage. It was being able to find her again.
Common questions
How do I find every video of my grandmother across all my tapes?
Find her on screen once, point at her face, and name her with a key. From that one confirmation the search looks for that same face across every tape — at other ages, angles, and cameras — and gathers what it finds into a single page for her, one best clip per tape. You name her once; it does the searching across the whole archive.
Does it use AI to guess who my grandparents are?
No — not for identity. Software finds where a face appears, but a person who knows the family confirms who it is. On a grainy VHS frame an automatic guess is often wrong, and one wrong guess can teach the system to confuse two relatives who look alike. Human-confirmed names are the reason the results hold up. An early automatic-matching experiment failed outright, which is why naming here is always done by a person.
Can I just do this myself with free software?
You can. Free, self-hosted tools will cluster faces on your own machine if you're comfortable running a server. The catch is that they guess identities, and on fifty-pixel VHS faces a guess quietly mislabels people and learns one relative from another. The done-for-you version trades the setup and the guessing for names a person confirms — and a sealed archive you never have to maintain.
What if my grandmother was always the one behind the camera?
Use the relationship graph. Once faces are named, the archive knows who was most often on screen with whom, so you can find a camera-shy grandparent through the people they were always beside — the holidays and trips where they'd have been in the room even when the lens rarely turned their way.