The Transcript Is the First Assembly Now

The first assembly used to be paper

On an interview-driven documentary, the first assembly is no longer a bin of clips. It is a transcript. That's a bigger change than it sounds. The first assembly is where a film decides what it is about.

The paper edit was the old version of the same idea. An editor reads the transcripts, picks the quotes and arranges them in narrative order, and each quote keeps a reference to its source and timecode, with notes for b-roll or music. It is a story told in other people's sentences, built before anyone touches a timeline. We have always liked the discipline of it. What we did not like was the labor around it.

What the transcript gives you

The old routine was printing timecoded transcripts, highlighting them by hand, and then going back into the footage to find each phrase. Building a string-out meant converting timestamps into clips, one at a time. No one's judgment is at work there.

Text-based editing removes that step. In Premiere Pro the transcript tracks playback, and text edits stay synchronized with the audio or video: select a phrase and it goes into a new sequence, highlight a passage and press Delete and the material comes out. The editor in that account builds a sequence per interview, pulls the sound bites, and assembles a radio edit, the audio-first version of the story, before any picture, music or graphics are added.

For us the radio edit is the useful artifact. You can argue about structure in text, where a move costs nothing, and listen to the result before anyone has committed to a shot. Arguments about order happen early, and that usually means fewer of the rounds of notes arrive late.

What selecting by word loses

A transcript is a flattened record. It keeps the words and drops nearly everything else. As one account of paper edits puts it, what looks good on paper doesn't always work as video: a quote can read well and fail on delivery, or the speaker's face can contradict what the sentence says. That's the weakness of the method, and it didn't go away when the transcript moved inside the editing software.

Selecting by word biases you toward the sentence that is cleanly constructed. People rarely say their truest thing in a clean sentence. The breath before an answer, the half-finished thought, the laugh that arrives a second late: none of it survives in the text, and all of it is performance. Even word search is a trap. It can find another take of a word you need to change a tense or an intonation, but it also finds instances that do not fit, so each result has to be checked by ear.

So we treat the transcript selects as a proposal, not a decision. Every select gets a pass with eyes and ears before it goes near the cut, and we keep a notes column for what the text cannot hold: the hesitation, the reaction, the moment someone looks away. It's also why the way we shoot interviews matters to the edit: the face has to be worth reading when the words are not enough.

A transcript is not a caption file

The accuracy bar for a paper-edit transcript is looser than the one for subtitles. A misspelled name or a stumbled sentence does not break a paper edit, because you are looking for story beats, not a verbatim record. That's a real convenience of automatic transcription. It's also a reason not to reuse the editing transcript as your subtitle source without a proper pass. Captions are a deliverable with their own standard, and a working transcript was never built to meet it.

Working note

Assemble in text, then sit with the footage. The transcript is a very good first draft of a film's argument and a poor judge of its feeling. For us the radio edit is also where the sound conversation starts, since it is the first time anyone hears the film's spine without picture to lean on. Build that version fast, and then go back to the performances and ask what the words left out.