How to prep a documentary for the edit: a field guide for new assistant editors

The job between the shoot and the cut

When people imagine documentary post, they imagine the edit — the year in a dark room where the film gets found. What they don't imagine is the job that makes that room possible: taking the raw chaos of a shoot day (three cameras, an external sound recorder, a field drive, a call sheet) and turning it into something an editor can open cold and understand.

That job is post-production prep, and if you're a new assistant editor, it is your job. This guide is the process I run on every non-fiction project — features, series, shorts. It's project-agnostic: the camera package, the drive names, and the project code change from film to film, but the sequence of steps and the reasons behind them do not.

One assumption before we start: I'll assume you're fluent in Premiere Pro, DaVinci Resolve, and multicam syncing. What I won't assume is that anyone has ever explained why the steps happen in this order — because the order is the whole system. Each phase preserves something the next one depends on. Above all, it preserves timecode.

(This is the manual craft, deliberately. If you want to see where AI now accelerates these same steps, I've written that up separately — Where AI actually belongs in documentary post-production. Learn this version first. You can't supervise an automation you couldn't run by hand.)

Five principles to internalize

Read these once, then never forget them. Every step in the guide is one of these principles wearing work clothes.

• Timecode is jammed before anything is shot. Frame-accurate timecode across every camera and the sound recorder is the backbone of the entire pipeline. Get it wrong on set and you pay for it in every phase after.
• Everything is organized by day, and mirrored. One folder per shoot day, dated so it sorts; camera and sound in their own subfolders; the same structure repeated for originals and proxies.
• Make your one working copy; the production keeps the redundancy. You copy the field drive to your originals drive. The offsite copies are the production's routine, not yours. Originals are irreplaceable — read-only the moment they're copied.
• Name it once, correctly, early. A clip is named right before it enters the edit, or it is never named right.
• Preserve the time-of-day gaps in the transcript version of the timeline. They're what lets anyone jump from a transcript or a rush straight to the frame in Premiere. (Phase 7 explains this — it's the most counterintuitive rule here, and the most important.)

The pipeline at a glance

Ten stops, one rule: don't reorder them. The transcript exists before the breakdown so you can break down against it. The gaps survive until the transcripts are exported. The reel names go on at transcode because of a conform that won't happen for another year. Prep is a long game played in the right order.

Phase 0 — Timecode before anything

Phase 0 happens before you receive a single file, but it defines everything you receive. On set, every camera — A, B, C, however many — and the external sound recorder are jammed to a common timecode, ideally frame-accurate.

Why it matters this much: when timecode agrees across sources, the day's multicam sync is instant and exact. And because every subsequence you later cut carries a timeline timecode that matches the source clips, a paragraph in a transcript, a burned-in timecode on a rush, and a spot on the Premiere timeline all point at the same frame. That unbroken chain starts here.

Your first task on every day of footage: even if you weren't on set, open the first clips of each camera and the sound and check that their timecode agrees. If it drifts or was never jammed, flag it immediately — it changes how you sync in Phase 4, and the team needs to know before the next shoot day.

Phase 1 — Back up before you touch a filename

The footage must exist in multiple safe places before you touch anything. Usually the material arrives as a field drive that was already offloaded and checksum-verified on set — you're receiving a verified copy, not making the first one.

Your job here is exactly one copy: field drive to your originals drive, made with a proper sync tool that verifies what it copied — never a drag-and-drop in the Finder. The broader redundancy — three copies encouraged, two the bare minimum, in different physical locations, sometimes LTO or cloud — lives in the production's own backup routine. Follow their protocol when they have one.

Two habits worth adopting early. First, modest drive sizes on purpose — I use 8 TB, deliberately not larger, so a rare drive failure loses a smaller chunk of the project. Second, and non-negotiable: once backed up, the originals are read-only. Everything downstream works off proxies. You will never have a reason to modify an original, so make it impossible to do by accident.

Phase 2 — Organize like the edit depends on it

Goal: a clean, correctly named, day-separated tree with zero naming collisions. Ideally the material arrives already split into one folder per shoot day; if it didn't, this is where you build that structure, and the proxies will mirror it exactly in the next phase.

Naming is where new assistant editors either build trust or quietly destroy it. The rule: front-load a sortable six-digit date, everywhere — on-disk folders, sequences, clips — so the drive and the Premiere project always line up. The rest of the token order can adapt per project; the rule doesn't.

Three tasks finish the phase. Rename what needs renaming: pro cameras embed sensible dated filenames; prosumer cameras (a GoPro, some mirrorless bodies) often don't — bring those clips to the project convention now, or they'll never match anything later. Hunt duplicates: in-camera counters reset between cards and days, so scan the whole tree for two files sharing a name and resolve every collision before proceeding — a duplicate basename can link the wrong clip in the edit or risk an overwrite. If anything is ambiguous, stop and ask; never guess. File the paperwork: sound and camera reports travel with the day untouched — you'll be glad they're there the day a take goes missing.

Phase 3 — Proxies, with the end of the film in mind

Transcode the camera originals to proxies in DaVinci Resolve — not Media Encoder — with Reel Names turned ON. This single setting is a message to your future self: the picture-lock timeline eventually comes back into Resolve for the conform and online, a year or more from now, and reel names written today are what make that conform reliable. It's the only reason Resolve is the transcode tool, and it's reason enough.

The spec: ProRes Proxy at 1920×1080 or close, chosen to preserve each camera's original aspect ratio — match the frame to the source rather than forcing 16:9. Proxies stay light on purpose; their only job is a responsive offline edit.

Ask the client one question before you batch anything: do they want clip name and timecode burned into the proxies, or clean proxies? If clean, the burn-in gets added later as an export effect on the rushes (Phase 7). Then mirror the day structure for the proxies, copy the external audio across, and verify: proxy counts must match originals per camera, and spot-check clips for full length and correct timecode. A truncated proxy won't announce itself — it'll fail mid-edit, weeks from now, as your problem.

Phase 4 — The sync map

Now the payoff for Phase 0. Each shoot day becomes one synced, multi-angle sequence — I call it the day's sync map — where every camera and the sound stay locked together.

If timecode was jammed correctly, this is a multicam sync by timecode with correctly labeled cameras: instant and frame-exact. If timecode is off, fall back to syncing by audio waveform; where even that fails, sync manually off the slate or by eye. Expect the fallbacks to be slower and less exact — which is precisely why you flagged the timecode problem back in Phase 0 instead of discovering it here.

The sync map files into its own stage of the bin structure (you'll see the full layout in Phase 6). From this point on, nobody should ever need to touch loose clips to look at the day — the sync map is the day.

Phase 5 — Transcripts and real names

Transcribe the source clips in Premiere — the external recorder audio is the cleanest source. If the recorder delivered a mix track plus isolated mics, transcribe from the mix for the offline unless the editor says otherwise.

Then the detective work. Premiere labels speakers generically and renumbers them per clip — the same person can be “Speaker 1” in one clip and “Speaker 2” in the next. Identify each real person using the call sheet, the transcript's own tells (“Speaker 2: My name is…”), and any director's notes. Then apply the real names to the source clips — not to a sequence — so the names ride with the footage into every timeline it ever appears in.

Done manually this is slow, honest work, clip by clip. Do it anyway, and do it now: every later phase — the breakdown, the transcripts, the editor's search for “the part where she talks about the verdict” — leans on these names being right.

Phase 6 — The breakdown is the judgment call

Break the day's sync map into scenes using Premiere's Make Subsequence — that function specifically, because a subsequence carries the timeline timecode, which matches the clip timecode in the sync map. The unbroken chain that began with the jam on set continues through every scene you cut here.

Where each scene begins and ends is your call — the one genuinely editorial decision in prep. Work from the transcript and the call sheet (the transcript was made in Phase 5 precisely so you'd have it now), and expect to split sections that run too coarse and merge ones split too fine, until the units match how the film will actually be discussed: this interview, that verité block, this B-roll run. The sections inherit the speaker names you applied in Phase 5.

Everything files into a bin structure that reads top to bottom in workflow order — raw, then synced, then the timelines you cut. The bins mirror the drive, so a clip's place on disk and its place in the project always match:

Follow any day through that chain — masters, sync map, flattened timeline, subsequences — and you'll find every version of its material. That traceability is the whole point of the numbering.

Phase 7 — Transcripts first, then rushes

Each day's timeline now leads two lives, and the order you produce them in matters more than anything else in this guide.

Export the transcripts first, from the version that still has the gaps. Time-of-day timecode naturally leaves holes in the timeline where the cameras weren't rolling. Keep them. As long as the gaps survive, the timeline's timecode still agrees with the burned-in timecode on the rushes and with the transcript — so when a director points at a paragraph, or reads a timecode off a review file, the editor types that timecode into Premiere and lands on the exact frame. Strip the gaps first and you've quietly destroyed that mapping for the whole day.

The transcript deliverables, per day: a per-scene transcript (one per subsequence, matching the breakdown one-to-one), a full-day transcript, and a short day summary — topics covered, who appears. Deliver as DOCX so the director can annotate.

Then build the video version — the rushes: multicam layout per the team's preference (each camera as its own file, a single wide, or a split-screen of all angles); dead video removed, except wild audio recorded intentionally without picture, which stays; external audio kept and everything else dropped — unless a section has only camera audio, which then stays; burn-in added now if the proxies are clean. Export light — low-res H.264, around 3 Mbps — built for fast review and upload, not for beauty.

Then QC every rush before it leaves. Sync, the right audio track, correct burn-in, correct layout, filename matching the breakdown. This pass is not optional. A bad rush costs the director's trust, and you only get to spend that once.

Deliver by the team's preferred channels — transcripts via Google Drive or Dropbox, rushes via Frame.io or Vimeo, sometimes a physical drive.

Phase 8 and hand-off — Archival and the drive guide

Archival runs parallel to the daily pipeline, usually alongside an archivist or visual researcher. Every archive piece gets an ID tag prefixed with the project's acronym, and the archivist keeps a master tracking spreadsheet — ID, source, rights holder, dates, licensing status. Your job is to make the media's ID match their sheet; theirs is the provenance record. Treat that record as seriously as the media itself: an unlicensed or unsourced clip cannot ship, no matter how good it looks in the cut.

Then the hand-off. Confirm the day is self-consistent everywhere it lives: originals, proxies, sync map and breakdown, transcripts, rushes — five locations, one day, no disagreements. And produce a drive and project guide: a short document explaining how the drive is organized and how the Premiere project is structured — folder tree, bin layout, naming conventions. It prevents the two things that actually break projects: renamed folders, and someone opening the wrong project file. Make one on every show, even when nobody asks.

Some gigs also request selects, stringouts, or a per-day shot log. Provide them when asked — but treat them as add-ons. They're editorial, not core prep.

Edge cases you'll actually meet

These come up often enough to know cold, and they share one rule: resolve them in conversation with the team; don't let them stall the core prep.

• Log footage. Leave proxies flat by default. A viewing LUT is applied only if the DP or director specifies one — and when they do, they usually provide it. Color-balancing footage is not the assistant editor's job unless asked.
• Multitrack audio. Ideal delivery is a mix track plus isolated mics. Editors differ on what they want synced — ask the editor or post supervisor. No answer available? Use the best sounding one for the offline.
• Drone. Just another camera. Give it its own letter, transcode it, treat it like any other angle.
• Stills and phone photos. Normally archival — unless they were shot specifically for the film, in which case they're their own category with their own handling.
• Vertical / social footage. Shot for social? Usually its own project entirely. Meant for the film? Its own camera with its own aspect ratio, resolved in the edit or online.
• Starting a new project. Copy a previous project and strip the bins you don't need — the skeleton already exists. Keep the numbered top-level order so the panel always reads raw → sync → cut. Better yet, make yourself a template project and copy it.

The definition of done

A shoot day is finished when every box below is checked — not before. Run it as a literal checklist until it's muscle memory.

That's the whole job. It isn't glamorous, and when it's done perfectly it's invisible — nobody ever compliments a sync map. But every documentary you've admired was carried, quietly, by someone who did this list right. If you're starting out as an assistant editor: learn the order, respect the timecode, and never guess when you can ask.

Questions about breaking into documentary post, or a project that needs this done? I'm easy to reach below. And when you're ready to see how AI accelerates this exact pipeline, read the companion post.