The DIT's second job: logging starts on set

Offload is half the DIT job, and it is the easy half. The other half is getting what you know about the footage out of your head before the drive back to the hotel deletes it.

The first card of the day came in at lunch. Copy, verify, manifest, eject, next. While it copied I knew things no file carries: the director bought take four but never said so with any conviction, the interview restarts at the third slate because a truck rolled through, the kid in the red jersey is the client's son and has to be in the cut. Three weeks from now an editor will need every one of those facts. By Thursday I will have lost most of them.

A verified copy answers exactly one question

The question is whether the bytes are safe, and good tooling answers it without your attention. Stream the file off the card with a running checksum, write it, read it back from the destination, hash it again, and log the whole exchange in an ASC-MHL manifest. Refuse to call the card wipeable until two destinations hold verified copies. Refuse to re-copy a card you already ingested. ClipLogger does all of this because machines are good at exactly this kind of diligence: no judgment required, only honesty at scale. And that is the ceiling of what a copy can do. The manifest can testify that every byte survived the van ride. It has no opinion about which of those bytes matter, whether the last setup was any good, or who the woman in the doorway is. A perfect copy of a mystery is still a mystery. You have only guaranteed the mystery two homes.

The gap between offload and edit is where knowledge dies

Every shoot day produces two assets: the media, and the understanding of the media. The media gets checksums, manifests, and redundant copies, an entire discipline built around never losing it. The understanding gets, on most productions, nothing, because at 2pm it feels permanent. It is not. It decays overnight, halves by the weekend, and is functionally gone by the time the project clears whatever stands between wrap and the edit: a funding gap, a re-slate, a different editor than anyone planned. Months later somebody opens the drive and finds A047C012.MXF four hundred times over. So they scrub, a full day at least, rebuilding a thin copy of what the person on set knew for free. I have been on both ends of that transaction. The set side barely notices the loss, because the knowledge leaks out quietly, one forgotten detail at a time. The edit side pays for all of it at once, on a deadline, at the hourly rate of whoever has to watch everything.

Write three things down while they are still true

You do not need my software for this part. You need a habit and a short vocabulary, and both are free. Pick three questions you can answer for every clip without deliberating: who is in frame, is the take usable, and what is it (scene, setup, moment, pick a word). Decide the allowed answers in the morning, not per clip, because "TD" and "touchdown" will otherwise fork into separate categories by dinner and someone will reconcile them in December. Then log at the moment of offload, while the card copies and you are stationary anyway, because that is the last window where the day is fresh and the clips are few. Do not save it for wrap; wrap-you is a stranger who remembers nothing. A text file beats memory. A camera report beats a text file. Anything written the same day beats everything written later. And if you keep only one habit from this piece, keep this one: flag the good takes before you sleep. That single bit per clip is worth more to an editor than a paragraph written next month.

Log at the speed of the copy

The copy takes the same minutes whether you supervise the progress bar or not, so that window is when the log happens. Once the first destination verifies, I open it in ClipLogger and work while the second copy trickles in the background: hover-scrub the thumbnails, flag takes, stamp fields from the keyboard, one key per value. With a roster loaded, the number is the name: hold shift, type 72, and it resolves to the athlete wearing 72, no typing names in a tent. On-device passes read the slates and transcribe the speech as clips open, so the interview is searchable before anyone thinks to ask for it. The AI proposes values against my own fields, and nothing is written until I confirm it, which keeps the log mine and keeps me looking at the footage instead of rubber-stamping a model's guess. Ten seconds a clip, spent inside a window that was already dead time.

Hand over a described card

The handoff is where the second job pays out. Confirmed values land in a plain JSON sidecar beside each file, plus whatever the pipeline reads: XMP for Premiere, CSV for Resolve, ALE for Avid. Rename-on-copy bakes the log into the filenames themselves, originals untouched, so the folder the editor receives says 72_TD_Johnson_A001C012.MOV where the camera said A001C012.MOV. The metadata follows the files wherever they move next; the sidecars still read in a plain text editor years from now. The difference downstream is not subtle. A mystery card starts with a day of scrubbing and a phone call to whoever was on set, if they remember, if they still work there. A described card starts with a search box and ends with a cut. Same footage, same checksums, entirely different first morning in the edit.

The manifest proves the footage survived the day; the log is the only proof the day itself did, and it can only be written on the day.