Last month a host emailed me at 11pm asking if we still had his raw files. His laptop had died, the SD card had been formatted for the next shoot, and the only copy of a 74 minute conversation he had spent three weeks scheduling was, in theory, sitting on a drive somewhere in his apartment. It was not.
We did have his files, because we keep everything we are handed until an episode is signed off and then some. But that is luck, not a backup strategy. His backup strategy was me.
I want to talk about this properly, because backup is the least glamorous part of running a show and the one that actually decides whether your show survives a bad week. Nobody posts about their folder structure. Everybody eventually pays for not having one.
The rule I give every host: three copies, or you do not have the footage
The old photography rule still holds. Three copies of every recording, on two different kinds of storage, with one of them somewhere else physically. If you have one copy on the camera card, you have zero copies. If you have the card plus a copy on your laptop, you have one and a half, because both of those live in the same bag and the same flood.
Here is what that looks like in practice for a normal show, and it costs almost nothing:
- Copy one: the working copy on your editing machine or a fast external SSD.
- Copy two: a cheap spinning hard drive that never leaves your desk and never gets used for anything else.
- Copy three: cloud storage, uploaded overnight, forgotten about.
The camera card is not one of the three. The card is the original, and the original gets protected until all three copies exist and have been checked. The moment you format a card before verifying three copies, you have decided your episode is disposable.
Format the card, and only after you have watched something back
The dangerous minute in any shoot is the one where you are packing up, the guest is putting on their coat, and you glance at a card that says 92 percent full. The temptation to clear it right there is enormous.
Do not. Copy first, verify second, format third, and format in the camera rather than on the computer so the file system matches what the camera expects.
Verify means actually opening the copied files and scrubbing through them. Not looking at the file size. Not trusting the progress bar. I have seen a 4K file copy across at exactly the right size and be unplayable past the eight minute mark because the drive was failing quietly. A file browser will tell you it is fine. Your eyes will tell you it is not.
If you record multi-cam, verify every angle, including the one you think you will not use. The angle you think you will not use is always the one with the only clean reaction shot, and if you are curious about how much that matters in the edit, I wrote about it when talking through proxy workflows for multi-cam.
Audio gets backed up separately, because audio is the show
Video is the packaging. Audio is the product. If your camera dies, you have a podcast. If your audio dies, you have nothing, and no amount of clever work in the edit builds a listenable episode out of a room mic recorded from four feet away.
So audio gets its own redundancy on top of the general backup.
- Record a safety track. Recorder plus camera scratch audio, or two recorders, or an interface plus a phone in the corner running voice memos. Something.
- Copy audio files first, before the video, every single time. They are small and they matter most.
- Keep the raw uncompressed audio even after the episode ships. Not the mixdown. The raw.
The reason for that last point is that six months later somebody wants a clip, or a sponsor wants a re-read swapped, or a platform wants a different loudness target, and a bounced stereo mix will not give you any of that cleanly. I have had to rebuild episodes from a mix, and it is grim work. If you want to see what is actually possible when the source is compromised, that is the whole subject of rescuing a badly recorded episode.
Backups are not about disaster. They are about the ordinary Tuesday when a drive stops mounting and you have a release scheduled for Thursday morning.
Name your files like a person who will be tired later
A backup you cannot search is a landfill. I have received drives with folders called "final", "final 2", "FINAL USE THIS", and "new folder", all containing partially overlapping footage from what turned out to be two different episodes.
The naming convention I ask hosts to use is boring and it works:
YYYY-MM-DD_EpisodeNumber_GuestOrTopic/ at the top, and inside it, folders for CAM_A, CAM_B, AUDIO, BROLL, EXPORTS.
Date first means everything sorts chronologically without any effort. Episode number means you can find it by the thing your audience knows it by. A short topic word means you can find it by the thing you remember. Three seconds of typing at ingest saves an hour of hunting in a year.
One more small thing that pays off constantly: put a plain text file in the episode folder with the recording date, who was on it, what gear was used, and anything that went wrong. Mic bumped at 22 minutes. Doorbell at 41. Guest asked to cut the bit about his old company. When you send that folder to an editor, or come back to it yourself in eight months, that file is worth more than any project file.
Cloud storage is fine. Cloud storage as your only copy is not.
I am not precious about which service. Use whatever you already pay for. What matters is that you understand what it is and is not doing.
Sync is not backup. If you delete a file on your machine and the folder syncs, it deletes everywhere. If your project gets corrupted, the corruption syncs. Versioning and trash retention buy you a window, usually 30 days, and that window is shorter than the average "wait, where did that go" discovery.
Also, upload times are real. A single 60 minute 4K multi-cam episode can be well over 100 GB. On a home connection that is an overnight job at best, so start it when you finish ingest and stop treating it as an afterthought you will do later. Later never happens on the week you actually need it.
For shows recording remotely with guests, get the guest's local recording off their machine within 48 hours. Not next week. People wipe their downloads folder, upgrade their laptop, or simply stop replying to emails once the excitement of being interviewed has passed.
What we keep on our side, and why that is not your backup either
Since we edit other people's shows, I should be straight about our end.
When you send us an episode, we hold the source files through the whole job and for a stretch after delivery, because revisions happen and because hosts come back asking for a clip from an old episode more often than you would expect. Human editors watch the entire episode here, so we tend to notice when a file is corrupt or an angle is missing before it becomes a problem, which is one of the quieter benefits of not letting software auto-cut anything by silence detection.
But we are a working copy, not an archive service. Our storage is part of a production pipeline. If your only copy of a two year old episode is on our drives, you have outsourced your show's survival to a company you do not control. I would say the same about any editor, including the good ones.
The version of this that works: you keep the archive, we keep the working set. When you want something old brought back, you send it over and we go again. That is also how our editing for solo shows is set up, and it is why our pricing is per episode rather than a storage subscription pretending to be an edit service.
Archive the project, not just the media
Once an episode ships, most people back up the raw footage and stop. Then a year later they want to change one line and discover that the edit itself is gone.
At sign off, collect and archive:
- The project file, consolidated so it only references media that exists.
- The final master export at full quality, not the compressed upload version.
- The raw audio stems if you have them.
- Any graphics, lower thirds, or intro assets used in that episode.
- The transcript.
Keep the transcript specifically because it makes your whole back catalogue searchable by what was said, which is how you find the clip you half remember. Just do not treat a machine transcript as gospel, for reasons I have gone on about at length in why I do not trust AI transcription alone.
Graphics matter more than people think. Rebuilding an intro to match an episode from eighteen months ago is a genuinely annoying afternoon. The same goes for the packaging around the episode. When our sister studio ClickTheory makes thumbnails for a show, we tell hosts to keep the layered files, not just the JPEGs, because a refresh later is ten minutes with the source and two hours without.
Test the restore, once, on a boring afternoon
Here is the part almost nobody does, and it is the only part that proves any of the above.
Pick an old episode. Pretend your main machine is gone. Restore that episode from your backup copies onto a different drive, open the project, and play it. Time yourself.
You will find something broken. Everyone does. A missing font, a media path that points at a drive you sold, an audio file that never made it because it was in a folder outside the project. Better to find it on a quiet Thursday than in the middle of a release week.
Once a year is enough. Twenty minutes. It is the cheapest insurance in this entire business.
You do not need a server rack. You need three copies, one of them off site, a naming convention, verified copies before any card gets formatted, audio treated as the priority it is, and one restore test a year.
Do that and a dead laptop becomes an annoying afternoon rather than a lost episode and a very awkward email to a guest who cleared two hours of his week for you.
If your archive is currently a pile of unlabelled drives and you want a hand working out what is worth keeping before you hand a season over for editing, write to me at samy@pixelsamy.com. I am in Dubai and we work with hosts everywhere, and I would much rather have that conversation before the drive fails than after.