How to build a rough cut from a transcript in Premiere

If you cut interviews, the rough cut is mostly a reading job. You listen, you log, you find the line that says it best, and you move it to the timeline. Premiere has had the tools to do that reading as text since 2021, and most editors still scrub. This is the transcript-first version, in two speeds: the built-in way that costs nothing, and the way we built Verbatim to do it from a brief.

Step 1: transcribe inside Premiere (free)

Premiere’s transcription runs on your machine and is included with your subscription. You can turn it on at import, or do it after the fact:

1. Select the clips in the Project panel (or open the sequence).
2. Window → Text, then Transcript → Transcribe.
3. Pick the language and whether to label speakers.

An hour of interview takes a few minutes. The result is a transcript with timecode on every word, which is what makes everything below possible. Do this for every interview before you start cutting; the whole point is to read the set, not one clip.

Step 2: the built-in way, select and Insert

Adobe added rough cutting from transcripts in 2022. In the Text panel’s Transcript tab, with a source clip open:

1. Search the transcript for a word or phrase. Premiere highlights every hit; click one and the playhead goes there.
2. Select the passage you want, a sentence or a paragraph, straight in the text.
3. Press Insert in the Source Monitor (or the comma key). The selection lands on the timeline as a clip.
4. Do the same for the next clip, and the next.

As you add clips, Premiere builds a sequence transcript. Switch the Transcript tab to Sequence view and you can read your cut top to bottom. To remove something, cut or delete the text and Premiere ripples the gap closed.

This works, and it is a big step up from scrubbing. It is also entirely manual: you find every line yourself, one clip’s transcript at a time, and nothing tells you that the line you want in interview three contradicts the line you took from interview one. On a two-interview piece that is fine. On a ten-interview documentary it is a week of reading before the first edit.

Step 3: the fast way, describe the story

Verbatim is a panel inside Premiere that reads the same transcripts and adds the step the Text panel is missing. Select every interview in the piece (timeline selection, project multi-select, or a sequence) and the transcripts open as tabs in the INPUT view, each speaker in their own colour, interviewer questions detected and listed.

From there, two routes to the timeline:

Pull selects. Type in FIND and the exact words match as you type. Or switch to AI search and describe the moment (“the part where she admits the first version failed”) and it finds the line even when none of those words were said. Click a line to take it, or drag across the words you want, and + Add to Timeline puts that quote on the output sequence as its own clip, cut at word boundaries.

Describe the story. Open Build Narrative and write a brief: who you want to hear from, the beats the piece has to hit, the energy, roughly how long. The editor proposes a cut across all the interviews and explains each choice. You reorder or drop beats, keep part of a quote, listen to any of them in the Source Monitor, and watch the running length change. Approve, and exactly that cut is built on the timeline, with each input clip’s Motion, Lumetri and audio settings carried over.

Every clip on the output sequence is then a story card. Drag to reorder and the timeline follows; delete and the gap closes; right-click a word to split. Verbatim’s own undo covers all of it, including a whole build. When the brief changes, Refine reopens the same conversation on the built cut and applies only the differences.

A few things that make either way go better

• Transcribe everything first, including the interviews you think you won’t use. The line you need is always in the one you skipped.
• Write the paper edit as a brief, not a list. “Open on the failure, let the engineer explain the fix, end on what it meant to the town” gets a better cut than a list of timecodes, from a person or from Verbatim.
• Give it a length. A target runtime is the single most useful constraint for a rough cut. Verbatim shows the running count as you edit the proposal.
• Keep the questions. Interviewer questions are the fastest map of an interview. Verbatim lists them; in the Text panel, label speakers so you can search for the interviewer’s name.
• Leave the footage where it is. None of this needs an upload. Premiere transcribes locally, and Verbatim sends only transcript text and your prompt to the AI you chose (or nothing, with the offline option: Ollama, on your Mac).

Where this does not help

Transcript-first cutting is for dialogue. If the piece is driven by visuals, B-roll logging and shot search are a different problem; see our comparison of AI rough cut tools for which ones do that. And if you are coming from Transcriptive, which was discontinued in May 2026, the transcription step above replaces the per-minute service, and Verbatim replaces the panel.

Try it on a real project: the interactive demo runs in your browser with a real podcast loaded, and the 14-day trial runs on your own interviews.