How to produce a feature-length YouTube documentary

Feature-length work is not a long episode. On September 14, 2026, YouTube announced that creator-led shows premiered at the Venice and Toronto film festivals, including Mark Vins' feature-length documentary Reef to Ridge, shot across the Galápagos. The same announcement notes the creators would bring their shows to YouTube later in the year — festival first, platform after. That sequence is one of the few things a creator team can copy directly.

The work itself is a production-management problem. Producing a feature-length documentary on a creator team means running a small number of explicit phase gates, keeping one owner per gate, standardising footage intake, and reviewing cuts in defined stages instead of one long thread. Everything else is craft, and craft is not what fails.

What follows is an operating model with illustrative numbers. Change the numbers to match your shoot; keep the gates.

Why the weekly pipeline breaks at feature length

A weekly channel operates on a short loop. An 18–22 minute episode, one or two shoot days, one editor, one reviewer, publish inside five to ten days. Ambiguity is survivable because the loop closes fast and the next upload corrects it.

Stretch that to 75 minutes and the loop stops correcting anything. Suppose you shoot 120 hours for a 75-minute film: that is 7,200 minutes of footage for 75 minutes of runtime, roughly a 96:1 ratio. Illustrative, but the direction is right for most documentary work, and it changes three things at once: how many people must agree on a cut, how long the gap is between shooting and publishing, and how many approval states a single cut passes through.

The failure mode is not a bad edit. It is a project that has no defined moment where the story stops changing.

Set four phase gates before the shoot

A gate is a decision with a named owner and a written exit criterion. Four are enough for a feature.

Assembly lock is the gate teams skip, and it is the expensive one. If your rough cut runs 95 minutes and your target is 75, you owe 20 minutes back. Decide now which scene leaves, not which seconds get trimmed across forty of them. Cutting 20 minutes by shaving frames is how a documentary loses its shape and its schedule.

Write the four gates on one page with dates and owner names, then put that page at the top of the project board as the first card. That single artefact prevents most of what goes wrong later. If your team already works to a stage-by-stage pipeline, these gates slot into it — see organising a video production pipeline.

Footage intake: one naming scheme, one log

Multi-location documentary shoots generate cards from several shooters on the same day. Without a convention, your editor spends the first week of the edit renaming files instead of cutting.

Pick a filename pattern and never deviate: PROJCODE-YYYYMMDD-CAM-CARD-CLIP — for example, REEF-20260314-A-004-0182. Project code, shoot date, camera letter, card number, clip number. Sortable, searchable, and stable across offloads.

Then keep one intake log with fixed columns: card ID, shoot day, location, shooter, offload operator, offload timestamp, checksum verified, backup destinations, proxy path, notes. Nothing gets a second copy until its row is complete. The wider version of this system is covered in organising raw footage from multiple shoots.

Storage is the intake decision that surprises people. Illustratively: 120 hours at about 100 GB per hour of camera original is 12,000 GB — 12 TB per copy. Three copies (working plus two backups) is roughly 36 TB before proxies and exports. Measure one card from your own camera rather than trusting a rule of thumb.

Documentary footage storage estimator

Size the working plus backup storage a feature-length shoot needs before the first card is offloaded.

Total storage equals hours shot multiplied by the camera's data rate per hour, multiplied by the number of copies you keep (working copy plus backups).

Assumptions: Illustrative inputs only. Camera data rates vary widely by codec and frame rate; proxies, project files and exports add more. Measure one card from your own camera before trusting any number here.

Worked example: Hours of footage shot: 120 hours; Camera data rate: 100 GB/hour; Total copies kept: 3 copies. Estimated storage needed: 36,000 GB.

Interview material deserves the same discipline as picture. Upload interview audio or video files to the workspace Brain and they are transcribed automatically, which gives you a searchable transcript for the paper edit and for fact-checking. Be clear about the boundary: kloudie can read that transcript and help you build an editing plan, but it cannot edit or render your footage, and there is no timeline editor to fall back on. See turning a recorded voiceover into an editing plan for where that handoff works well.

Assign four owners, not four departments

Feature-length projects stall at handoffs, not at tasks. Name a person for each of these, even if one person holds two on a small team:

• Story owner — final call on what the film is about, and the only person who can reopen story lock.
• Field production owner — shoot schedule, releases on location, card handling.
• Media manager — intake log, backups, proxy generation, archive integrity.
• Post producer — cut schedule, review rounds, finishing list, delivery.

Then set access so each person sees only what their role needs. Crew and freelance editors are well served by the guest tier without the view-all-cards capability: card visibility is enforced server-side, so an editor without that capability sees only the cards they are assigned to, and an unassigned card stays invisible. Internal assembly cards, budget notes, and unfinished interview selects belong on restricted boards rather than a shared one.

The full permission setup is in giving project access without exposing internal work, and the guest-editor version is covered in scoping freelance editors to assigned projects.

Review in stages, not in one long thread

Feature review fails when every reviewer comments on every cut. Fix it by defining who is allowed to say what at each stage.

Two rules make this work. First, one person collects and consolidates notes per stage; individual reviewers do not submit parallel lists. Second, every note names a timecode. "The middle drags" is a mood; "05:12–06:40 feels long, consider ending on the door closing" is actionable.

Version history is what keeps stage confusion out of the review. Upload a new cut as a new version of the same file rather than as a new file, and reviewers can switch between versions while the comments stay attached to the work. Timestamped comments and @mentions on video and documents are the mechanism; the process around them is documented in the file-version approval workflow and the video review and approval process.

The common failure: internal assembly and external review living on the same card. A stakeholder then reviews a cut that was replaced two days earlier. Keep the internal string-out on a restricted board and share only the cut you want reviewed.

Rights, music, and the deliverables calendar

Feature delivery is blocked by paperwork more often than by picture. Keep one list, visible to the post producer, with a state per item: contributor releases, location releases, archive and third-party footage licences, drone or location permits, composer agreements, and a music cue sheet with composer, publisher and performing rights details.

Add one line to that list: distribution sequence. The festival-first, YouTube-after pattern in the September 2026 announcement is a sequence, not a rule. Festival eligibility, premiere requirements, and online-publication rules differ between festivals and change between cycles. Check the current regulations of each festival you intend to enter before you publish anything, and record the date you checked.

Then build the deliverables calendar backwards from picture lock. Offsets below are a working example for a ~75-minute film on a small team. Adapt them; the point is that each item has a date before picture lock happens, not after.

• Picture lock, day 0. Finishing list frozen.
• Day +3: master files and any festival-screening deliverables exported and QC'd.
• Day +7: captions and subtitles delivered, transcript reviewed against the final cut.
• Day +10: title, description, chapters, thumbnail options, end screen.
• Day +14: publish window opens. Publishing requires a supported connected account and the appropriate workspace permission, so confirm connection status before you commit to a date.
• Day +21: clip pack and follow-up posts from the film's strongest scenes.

If a festival premiere comes first, shift the publish window rather than compressing finishing. Finishing under a squeezed calendar is where a year of fieldwork turns into an audio problem nobody catches until the premiere.

The week-one deliverable

Before the next shoot day, produce one page: the four gates, a named owner for each, target runtime, and the date you intend to publish. Put it at the top of the board as a card, and add the intake log as the second card with its columns already defined.

Then run one test: assign the media manager a real card, keep them scoped, and confirm they see only that card and the Drive folder they need. If you are still deciding who counts as a payable seat while adding crew, the tradeoffs are in free client seats versus paid user seats, and current plan details are on the pricing page.

A feature-length documentary from a creator team is achievable. It just needs a structure that survives the months between the first shoot day and the locked cut — and that structure is four gates, four owners, one naming scheme, and a review process where each stage has a single consolidating voice.