How to cut a long recording to an exact written shot list

Last updated 2026-08-17 · 6 min read

What is the short version?

Sometimes you already know the edit. You have watched the recording, chosen the passages and written them down — and everything that remains is executing that list exactly, which is mechanical work that still costs an entire evening to do properly.

1. Go to Vizard Agent and upload the recording.
2. Paste the cut list: the passages you want, in order.
3. Say the shape, the target length and the subtitle style.

What do you need before you start?

The recording and your list. Vizard Agent finds the exact word boundaries for each passage you name, so your list can quote the lines rather than give timecodes — which is far easier to write and far harder to get wrong.

• The recording. One long take. Length is not a constraint.
• The cut list. Quoted lines, timecodes, or a mixture. Both work.
• The order. Say if the reel's order differs from the recording's.
• The shape. Vertical from a wide recording means a crop, and the crop needs checking.
• The subtitle style. These reels are usually watched muted.

What do you type into Vizard Agent?

Paste your list as written. Vizard Agent treats a written cut list as a specification to be executed exactly rather than as a suggestion, which is precisely what you want when you have already made the editorial decisions yourself.

Prompt

Variants worth knowing:

• Quoted lines instead of timecodes. Easier to write, and Vizard Agent finds them in the transcript.
• A tolerance. Say whether a passage may be trimmed if the total runs long.
• Keep the original captions off. If the source has burned-in text, say so before the crop is chosen.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real scripted cut. The first thing it does after looking at the footage is test the crop — because a vertical reel from a wide recording is a framing decision that has to be verified before any cutting is worth doing.

1. Analyses the recording's format, resolution and duration.
2. Extracts frames at several points to see the composition and where the subject sits.
3. Views the frames one by one to check the subject stays in the same part of the picture.
4. Runs a test vertical crop centred on the subject and looks at it to confirm nothing is cut off.
5. Checks two more frames to confirm the framing holds later in the recording.
6. Transcribes the whole recording and inspects the word-timing structure in code.
7. Extracts the exact word timings for every passage in the list, in batches across the whole file.
8. Maps and saves all the timings so each cut point is a measured number rather than an estimate.
9. Builds a re-aligned subtitle file against the concatenated timeline, so the captions match the assembled reel rather than the original recording.

Step 9 is the step people forget. Once passages from 2:10, 4:20 and 7:05 are joined, every subtitle timing from the original file is wrong — and rebuilding them against the new timeline is the only way the captions match.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 30fps, 110.95 seconds, AAC audio. Vertical, just under two minutes, cut exactly to the written list with the subject framed centrally and subtitles aligned to the assembled timeline.

Vizard Agent verified the crop on four separate frames before cutting anything, which is what stops a vertical reel from beheading its subject halfway through.

Vizard Agent keeps the full word map from the transcription, which is why a later change to one passage's boundary costs a re-assembly rather than another pass over the whole recording from the beginning.

Timing was not measured separately for this kind of job. The closest measured work runs a median of 28 to 38 minutes end to end, with the middle half spread considerably wider. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.

When does this not work well?

An exact list is executed exactly, which is the point and also the risk. Vizard Agent does what the list says and verifies the mechanics around it, and the list itself is where all the editorial judgement lives — so an error in your notes becomes an error in the finished reel.

• A wrong timecode is cut wrongly. Quoting the lines is safer than trusting your notes.
• The crop may not hold. If the subject moves across a wide frame, no single vertical crop works throughout.
• Passages assume their context. A line that made sense at 4:20 may not stand alone at 0:30.
• Length and completeness conflict. A full cut list very often runs longer than the target length; say plainly which of the two should give way.
• Burned-in captions travel. Source text in the picture stays inside the crop unless the framing deliberately avoids it.
• A list is not a structure. Passages chosen for their content still need an order that builds, and that ordering is yours.

How do you fix a result that came back wrong?

Name the passage. Vizard Agent keeps the full transcript, every mapped word timing, the tested crop and the re-aligned subtitle file, so changing a boundary or dropping a section is a re-assembly rather than another pass over the recording.

• "Start that passage a line earlier." A new boundary from the same word map.
• "Drop the third section." Removed and the subtitles re-aligned.
• "The crop cuts off his hand." A wider framing, tested on frames again.

How does Vizard Agent compare to doing it yourself?

By hand this is scrubbing a long recording to find each quoted line, setting a dozen in and out points by eye, cropping the whole thing vertically, and then re-timing every single subtitle because joining the passages moved all of them.

Common questions

Can I give quoted lines instead of timecodes? Yes, and it is safer. Vizard Agent finds them in the transcript and cuts at the word.

Will it follow my list exactly? Yes. A written cut list is treated as a specification rather than a suggestion.

Will the subject stay in frame? Vizard Agent tests the vertical crop on several frames before cutting and will tell you if the framing cannot hold.

Do the subtitles match after assembly? Yes. They are rebuilt against the concatenated timeline rather than reused from the source.

How long can the recording be? Any length. Vizard Agent transcribes it once and works from the timings.

What if my list runs over the target length? Say whether to trim passages or extend the reel; Vizard Agent will follow whichever you choose.

Can it reorder the passages? Yes. Vizard Agent cuts in whatever order your list specifies, not the recording's.

Can it add music underneath? Yes. Vizard Agent mixes a track under the speech and ducks it where the words land.