Paper Edit

A paper edit is a plan for a cut written from transcripts before any footage is assembled: a document listing which lines, from which takes, in which order, with timecodes so an editor can find them. The practice comes from documentary and interview-driven filmmaking, where producers and editors cut the story on paper first because structure is easier to see in text than in hours of footage. In 2026 the term returned as a software feature when Adobe Premiere 26.5 shipped a tool called Paper Edit. Transcript-driven AI editing has also made the paper edit the natural format for an agent's proposed cut.

How a Traditional Paper Edit Works

  1. Transcribe everything with timecodes and speaker names, so every line can be traced to a source clip and moment.
  2. Mark selects. Read the transcripts and highlight the strongest lines, the usable answers, the moments of emotion or clarity.
  3. Build the structure. Arrange selects into story beats: an opening, a problem, turning points, an ending. Rearrange freely, because moving a paragraph costs nothing.
  4. Annotate. Note cutaways and B-roll ideas, flag lines that need a cleaner take, and mark where music or a pause should breathe.
  5. Assemble and test. The editor builds the sequence from the document, often first as a "radio cut" of sound alone, and the team watches to see what works on screen rather than on the page.

The paper edit is then revised against the assembly. Lines that read well and play badly come out, and moments with no words, such as a look or a laugh, go in.

Why the Practice Survived

Paper edits have outlasted every generation of editing software because they solve problems that timelines do not:

  • Structure is visible. A 90-minute interview reduced to two pages of selects shows the shape of a story in a way a timeline cannot.
  • Collaboration is easy. Directors, producers, clients and subject-matter reviewers can all read and comment on a document. Few of them can work a nonlinear editor.
  • Review of what is said. Accuracy, legal and compliance checks apply to words, and a paper edit puts the words in front of the reviewer.
  • Iteration is cheap. Trying three structures in a document takes minutes. Building three sequences takes hours.

From Practice to Feature

Software has absorbed the paper edit in stages. Transcript-based editors such as Descript made the transcript itself the editing surface (see transcript-based editing). DaVinci Resolve 20 added IntelliScript, which builds a timeline from a written script. Adobe Premiere 26.5 (September 2026) went furthest in name. Adobe describes Paper Edit as letting editors "select the sentences or dialogue you want to keep, and preview your selections before creating a new sequence," built "for interviews, podcasts, and dialogue-heavy edits where the transcript is your blueprint." In Adobe's framing, "Your edit plan is already written—now you can build it in seconds."

These features remove the most tedious step, finding each selected line in the footage and laying it down. They do not replace the judgment of choosing lines, which is still the human part.

The Paper Edit as an Agent's Plan

A paper edit is a human-readable description of a cut, which makes it a good format for agentic video editing. An AI agent reading transcripts can propose a structure as a paper edit, with each line quoted, its source and timecode given, and a short reason for its place. A person can review, reorder and strike lines in the document before anything is rendered. That review-then-apply flow is the plan-and-apply pattern, and it keeps the most consequential decision, what the video says, in a form anyone can check.

Taken further, the paper edit becomes a lossless text view of the edit itself: a cut sheet that can be edited and turned back into a frame-accurate timeline. Plain text such as Markdown suits this, because it can be versioned, diffed and reviewed with the same tools used for docs as code. This is one face of video as code.

Writing a Good Paper Edit

  • Always include source and timecode for every line, so assembly is mechanical and nothing is lost.
  • Label speakers. Accurate speaker labels prevent the most embarrassing errors.
  • Correct the transcript first. Misheard names and words carry straight into captions and search. Speech recognition is good but not perfect.
  • Note what is not said. Mark reactions, silences and visual moments worth keeping, since the transcript cannot show them.
  • Estimate length from the time ranges, not the word count, because pauses are part of the runtime.
  • Watch the assembly. Reading is for structure. Watching decides what stays.

Limits

The paper edit's blind spots are those of text. It cannot judge performance, rhythm or the picture, and it favors talkers over moments. It works best for interviews, documentaries, podcasts and explainers, and worst for action, music-driven or visual storytelling. That is where a timeline-first approach still wins. For the wider set of AI editing techniques, see AI video editing.

Further Reading