Camera media cleanup is mostly straightforward until it isn't. The "isn't" is a small set of edge cases (XML sidecars, spanned clip segments, multicam source clips) that account for almost every horror story I've heard about a cleanup pass going wrong. These are the cases where deleting a file that looks safe quietly destroys hours of editorial work.

If you've ever been afraid to delete anything from a camera card folder, this is why.

XML Sidecars: The Files Editors Forget Exist

Most modern cameras drop more than one file per clip. Sony, Canon, Panasonic, and ARRI all generate XML sidecars alongside the actual video file. Those sidecars hold timecode, lens metadata, color science settings, audio routing, focus distance, and increasingly, look LUT references.

Think of the XML sidecar as the slate, the camera report, and the script supervisor's notes all stapled together. The video file is the take. Throw away the staple-stack and the take still plays, but every piece of context the camera department captured is gone. Audio sync against scene/take numbers? Broken. Lens metadata for VFX? Gone. Color decisions baked at the camera? Defaulted.

The cleanup gotcha is that most editors only think about the video files. A script that deletes .MXF or .MP4 and ignores *.XML leaves you with sidecars pointing at clips that no longer exist. A script that deletes the XML and keeps the video silently strips the metadata. Either direction breaks something.

The rule is simple even if the implementation isn't. The sidecar and its clip are one unit. They live together, they die together.

Spanned Clips: One Take, Many Files

A long take from a RED, ARRI, or Sony camera doesn't fit in a single file. Cameras automatically split it at a size threshold (often 4 GB for FAT compatibility, sometimes higher) and produce a sequence of segment files: _001, _002, _003, and so on. To Premiere, that's one clip. To your file system, it's three or four files with consecutive names.

Here's where it gets dangerous. If your cleanup logic checks files individually instead of as a span family, you can end up in any of these failure modes:

The reference chain has to be walked at the segment level, not the clip level, before any spanned camera media gets touched.

Multicam Source Clips: One Edit, Many Originals

Multicam takes the gotchas above and stacks them. A multicam clip in Premiere is a virtual sequence built from two, four, eight, or sixteen individual source clips synced by timecode or audio. When you cut to the multicam clip on a timeline, you're not cutting one source. You're cutting all of them, simultaneously, even though only one camera's video is showing at any given frame.

The cleanup trap is obvious in retrospect. An editor sees the program output uses Camera 2 for the entire take, and reasons that Cameras 1, 3, and 4 are unused. They aren't. If anyone ever opens that multicam clip again to switch angles, the missing sources become media-offline immediately. The multicam source clip is only "in use" if all of its constituent angles are present.

This means a usage check that walks from sequences down to MasterClips has to recognize multicam source clips as a special case: a single MasterClip reference that pulls in N actual files from disk. Treat any of those N files as orphaned and you've broken the multicam.

Layer in spanned segments and XML sidecars on top of that (a four-camera shoot where each angle is a spanned RED take with sidecars) and one multicam clip can easily reference 30+ files on disk. All of them have to be present, all of them have to travel together, and a cleanup pass that misses any one of them silently breaks the clip.

What a Safe Pass Actually Looks Like

The pattern that survives all three gotchas:

  1. Group files by clip family before scanning. A clip is its video file plus its sidecar plus all its spanned segments plus, in multicam cases, all its sibling angles. Treat the family as the atomic unit.
  2. Walk every project on the drive, not one. Single-project analysis lies to you when projects share media.
  3. Resolve multicam source clips down to their constituent files during the reference-chain walk. The multicam parent is in use if and only if every angle is reachable.
  4. Never delete one member of a family without the others. Either the whole clip family is unused everywhere, or none of it gets touched.

This is unglamorous work that nobody wants to do by hand, which is exactly why almost nobody does it, which is exactly why drives keep filling up with footage that's too scary to clean.

Reclaim Math

The formats that produce these gotchas are the heaviest formats on your drive. Weighted average for a typical multicam doc shoot (RED A-cam at 1.5 GB/min, BRAW B-cam at 800 MB/min, ProRes 422 HQ C/D-cams at 600 MB/min) lands somewhere around 850 MB/min per camera, times four cameras, times whatever percentage of those takes never made any cut.

Run the real numbers for your own drive in the calculator before deciding whether the manual approach is worth your weekend.

A Tool That Handles the Family

Clip Sweeper groups clip families before it scans, walks multicam source clips down to every constituent angle, and treats spanned segments as atomic units, so the gotchas above don't become silent breakage. Worth pointing at a multicam-heavy drive and seeing what's actually safe to remove.