Try it now
A working sandbox. No sign-up, no project. Sample data only.
- Rows with speaker, in and out, source file and verbatim text. Drag to reorder, type act headers inline.
- Play it back as a radio cut, here or on a phone.
- Three readings of the cut's shape: acts as cards, pace over running time, who holds the floor.
- Versions of the whole cut; a pass on each row saying which round it arrived in.
- Notes, story turns flagged on the row, cards for what the cut still needs, scratch narration.
- Import from Word, Excel or PDF, or from highlighted Word transcripts.
| IN | SPEAKER | TEXT | |
|---|---|---|---|
| 00:04:31 | ELENA M. | It wasn't planned. None of it was. | |
| 01:12:04 | DAVID R. | We had run the simulation a hundred times. | |
| 02:44:11 | MARCUS T. | I followed protocol. That's all I'll say. |
How it works
Three steps from raw material to result.
Take me back to that night.
It wasn't planned. None of it was. We were sitting there… and the lights went out.
Highlight a line in any transcript and send it to the paper cut. The row lands with speaker, timecodes, and source file linked, drawn where you put it in the same frame.
| IN | SPEAKER | TEXT | |
|---|---|---|---|
| 01:12:04 | DAVID R. | We ran the simulation a hundred times. | |
| 00:04:31 | ELENA M. | It wasn't planned. None of it was. | |
| 02:44:11 | MARCUS T. | I followed protocol. |
Reorder rows by drag, type act headers, and trim verbose selects. The AI assist can regroup and propose headers. Flag the moment the story turns on the row where it turns.
Step back into the board, the pace curve or the banded view to see where the running time actually sits. Then send the package to the edit bay.
Hear it before the timeline
Press play and the paper cut plays as a radio cut. Click a word to jump there; drag the playhead to scrub.
The shape of the cut
Three readings of the same numbers: acts as cards, pace over running time, who holds the floor.
Acts across, beats down. Where the running time sits.
Frequently asked questions
What exactly is a paper cut?
A written assembly of interview selects in the order you want them in the edit, with speaker labels, in/out timecodes, and the text of each quote. Editors call it a paper cut, a radio cut, or a paper edit depending on the tradition. The structure is the same.
How do I add a select to the paper cut?
Highlight the line in any transcript and use the "Send to paper cut" action. The select lands as a new row with speaker, timecodes, and source file linked.
What do the three shape views tell me?
The board lays acts across and beats down, so you can see that act two is carrying twice the running time of act three. The pace curve puts running time on one axis and how finely each stretch is cut on the other, which is where you spot a four-minute answer sitting in the middle of a fast sequence. The banded view gives each act a lane coloured by voice, so a lane held by one person reads as a slab and a real exchange reads as a weave. All three are computed from the same numbers as the Duration column, so they cannot disagree with the sheet.
Why is search in a paper cut typo-tolerant?
Because the words in a cut were transcribed rather than typed. You are searching for what somebody said, through a machine's guess at how they said it, from your own memory of hearing it. "Alameda" comes back as "Alamedia" and a name is spelled three ways across four interviews, so an exact match finds none of it. Terms do not have to be adjacent or in order either, because people search a cut with two or three remembered words.
Can I leave notes on a row?
Yes, and on transcript passages and beats as well. Name a colleague in the note and it lands in their notifications with a link back to that row. Anyone who can open the document can leave a note, including viewers. Marking one resolved is an editorial call and needs edit rights.
Can I import an existing paper cut from Word or Excel?
Yes. The import wizard reads .docx, .xlsx, and .pdf paper cuts and maps each row to a select, attempting to auto-link to the source transcripts where it can.
Does the AI assistant write text for me?
It helps with reorganising, suggesting section headers, and trimming verbose selects. It does not rewrite a speaker's quotes. Those stay verbatim.
What is the difference between a version and a pass?
A version is a copy of the whole cut at a moment, kept so you can go back to it. A pass is per row: it records which round of the cut a bite arrived in, so a producer reading the sheet can tell the first assembly from what was added after the director's notes.
What is the synthetic voice for?
Scratch narration. A line of voice-over that has not been recorded yet can be read by a temporary synthetic voice so the assembly plays end to end, or you can record it yourself in the browser and have the noise cleaned up. Generated audio is flagged as generated, the feature is one of the project's AI switches, and it never replaces anything a subject said on tape.
Can I listen to the cut on my phone?
Yes. The app installs to a phone's home screen and the paper cut plays back there as a radio cut, so a director can hear the assembly on the drive home without an export.
Related capabilities
Further reading
Background guides and comparisons.
Three terms from different production traditions describe the same practice: assemble interview selects in writing before touching the NLE. Here is where each term comes from and when the distinction actually matters.
Descript's text-based timeline editing and a spreadsheet paper cut both start from transcript text, but serve different moments in the workflow. Descript is a nonlinear editing environment; a paper cut is a plan.