Documentary remote review:
the cut travels, the drives stay home.

The drive habit made sense when there was no other way to move a long cut. It survives because it feels safe, and because the first cost is invisible. The second cost is not.

A drive arrives and the reviewer plays it in whatever player is on their machine. You have no idea which decoder they used, what their display is doing to your grade, or whether they watched the file you sent or an older one still sitting on their desktop. Their notes come back as a paragraph in an email, or as a phone call, and somebody in the cutting room translates a sentence about the bit after the archive into a frame number by guessing. Multiply that by four reviewers and a year.

Then the round repeats. A new drive, another courier, another week of transit into a schedule that had no week to give. Meanwhile the version question gets worse: three people hold three different physical copies, each labelled with a date, and none of the labels tells you which one was approved. When a broadcaster asks in March which cut legal signed off in November, the honest answer is that nobody knows, and somebody spends a day reconstructing it from an inbox.

The fix is not more discipline in naming files. It is to stop moving copies at all. One room per film, one link, versions stacked inside it, and every decision recorded against a specific version rather than against a memory.

Two habits make this hold up over a long schedule. Name versions after the stage rather than the date, because assembly, rough cut, fine cut and picture lock mean something to a funder and 09_11_v4b does not. And close each round explicitly by working the note list to done, so that the next reviewer opening the room sees an empty list rather than a year of accumulated opinion.

The instinct on a remote documentary is to schedule a call. A director in Berlin, a producer in Montreal and an EP in Los Angeles can find one overlapping hour, and in that hour everyone watches together and talks. It works once. It does not survive a year, because the overlapping hour is expensive, because a ninety-minute cut does not fit in it, and because the reviewer who could not make it reviews nothing.

The alternative is not worse, it is better, provided the notes carry enough structure to stand on their own. A timecoded note with a drawing on the frame is more precise than a sentence spoken over playback, because it names the moment exactly and it survives being read three days later. A per-version approval is more useful than a verbal yes, because it is attached to a file. The call then becomes what it should be: a short conversation about the two disagreements that the written notes exposed, rather than a ninety-minute group screening on three different connections.

One practical note on bandwidth. Views, uploads and viewers are not metered here on any plan, including the free one, so a reviewer in another country rewatching a reel four times costs nothing and triggers nothing. Plans have storage, seat and caption allowances; streaming has no extra charge. That matters more than it sounds on a project where the same cut is watched by a dozen people over a month. The timecoded comments guide covers what a note stores and how it reaches your timeline as a marker.

A documentary carries reviewers who are not judging the film. The archive producer needs the four shots that came from one collection. The lawyer needs every sentence where a named person is accused of something. The fact checker needs the claim about the second inquiry. On a ninety-minute cut, each of those passes is a scrubbing exercise that eats a day.

Captions change the arithmetic. Once a film has them, generated by a speech to text model, the dialogue is searchable: type a phrase and every moment it is spoken becomes a button that seeks the player there. The lawyer searches a surname and reads the four passages instead of watching the film. The fact checker jumps straight to the claim. The translator working on subtitles reads the cue list rather than transcribing from scratch. Details of what that search does and does not find are in the dialogue search guide.

The limit is worth stating plainly. Search finds spoken words. It does not find a shot, a face, a piece of music or a location, and it cannot tell you which archive collection a clip came from. Those still live in your own log. What it removes is the specific waste of a reviewer hunting for a line they already know is in there somewhere.

The pattern that suits a long project is to keep the live reel and the current cut hot, and push superseded versions and finished selects to cold once a stage is signed off. A cold master does not stream, embed or review, so it is storage rather than access; bring back only what you are about to open. Archiving frees hot space immediately, and restores are size-verified.

There is no rendition ladder, which is the direct consequence of not transcoding. On a weak connection the player buffers rather than dropping to a smaller version, because a smaller version was never made. A reviewer on a train will have a better time with a tool that builds proxies. The practical answer on a long cut is to send a lighter export when reach matters more than the data rate, and to keep the high-bitrate export for the people judging the grade.

Camera formats are not review formats. Browser playback is H.264, HEVC and AV1, and 8K plays when the viewer’s own hardware decodes it. ProRes, ProRes 4444 and DNxHR are hosted for download rather than streamed, so a documentary finishing in ProRes needs a viewing export for the round and keeps the master for the hand-off.

And if your editorial team already lives inside one hub for dailies, bins, assistant work and review, moving only the client-facing half of that out is a real disruption for a modest gain. The case for changing is strongest when the people reviewing are outside your company and the project runs long. The review and approval pillar has the general checklist, and the hot and cold storage guide covers what keeping a finished film online actually costs.