Format guide
Turn forty hours of footage into a story worth ninety minutes
Documentary editing is triage before it is craft: interviews, verite, and archive arrive as a mountain of maybe. This guide covers logging, selects, transcript-based structure, and the B-roll and sound work that turns research into a watchable film.
From the Mellius product team · Updated September 28, 2026
Conspiracy theorists, you may hate this
AI Director
›
Just type what you need
- ›Remove all silent parts
- ›Add bold kinetic captions
- ›Only have Bill Ackman speaking
- ›Cut to highlights only
- ›Add cinematic transitions
- ›Match cuts to the beat
The short answer
Log and transcribe everything, then pull selects organized by theme rather than shoot order. Build a paper edit from the strongest soundbites, structure three acts around a question the film answers, cover jumps with motivated B-roll and archive, and mix dialogue first so every word stays intelligible under music.
Honest structure
The edit clarifies what happened without manufacturing quotes or reversing cause and effect.
Audible testimony
Interview dialogue stays intelligible and well-paced, with trims that preserve each speaker's meaning.
Motivated coverage
Every B-roll shot illustrates, locates, or emotionalizes something the audience just heard.
Log, transcribe, and build the paper edit
Ingest everything with consistent bin names — interviews, verite, archive, drone, stills — and transcribe all speech before judging a single frame. Rate soundbites on a simple scale and tag them by theme, not by shoot day, so selects for topics like childhood, conflict, and resolution pull together regardless of when they were recorded. Most features log at ratios of twenty to fifty hours of footage per finished hour, which is why searchable selects beat heroic memory.
Write the paper edit as a sequence of bites on paper or cards before touching the timeline: opening question, act turns, and closing answer, each supported by two or three bites that genuinely connect. Cut the radio edit first — audio only, no picture — and listen for whether the story holds without any visuals. If the spine works blind, the picture edit becomes illustration; if it does not, no montage will save it.
- Transcribe before selecting; untranscribed footage gets forgotten.
- Tag selects by theme so storylines assemble across shoot days.
- Cut a radio-only assembly and test the story without picture.
- Keep full-transcript context one click away for every bite used.
Cover, pace, and mix for comprehension
Cover interview jump cuts with B-roll that proves or places the claim: the street named, the hands described, the document quoted. Vary rhythm deliberately — dense montage for context, long verite holds for emotion — and let silences and room tone breathe rather than crossfading every pause into oblivion. Archive and stills need motion restraint: a slow push over a photograph reads as care, while constant fast zooms read as panic.
Mix dialogue first with levels riding between -12 and -6 dB on peaks, then duck music six to ten decibels under speech so narration never fights the score. Keep one continuous room-tone bed under each interview scene to hide edit points, and fact-check every on-screen date, name spelling, and lower-third title before lock. In Mellius, you can ask to remove filler pauses or rebalance music under dialogue across a whole scene, then undo any single change since the timeline stays fully editable.
A practical sequence
Put it into practice
- 01
Organize the mountain
Log, transcribe, and tag every clip by theme so selects assemble by storyline, not shoot order.
- 02
Cut the radio edit
Structure the strongest bites into three acts and confirm the story works with audio alone.
- 03
Cover and finish
Lay motivated B-roll over jumps, pace montage against verite, and mix dialogue above everything.
Frequently asked questions
How do documentary editors organize so much footage?
They transcribe everything and tag selects by theme rather than by shoot day, so all bites about one storyline sit together. Bins separate interviews, verite, archive, and stills, and a paper edit sequences the strongest bites before picture is cut. A radio-only assembly then proves the story works before any finishing begins.
How do you shorten an interview without changing its meaning?
Cut filler words, false starts, and repeated phrasings while keeping the sentence the speaker actually landed on. Cover the resulting jump cuts with B-roll that illustrates the point being made. When in doubt, keep the longer version — audiences forgive length more readily than they forgive feeling manipulated.
What should documentary B-roll show?
The best B-roll proves, places, or emotionalizes what was just said: the location named, the process described, or the face reacting. Avoid generic stock that contradicts the testimony's specifics, like wrong-season exteriors or mismatched workplaces. Every shot should answer why it appears at exactly that moment.
Continue reading
Primary sources
Product capabilities change. These official references support the comparisons above.
Describe the edit. Keep the timeline.
Mellius turns conversational direction into cuts, captions, graphics, motion, color, and sound you can still inspect and change.
Start editing free