How to cut a long product roundup into short clips

Last updated 2026-08-16 · 5 min read

What is the short version?

A long product roundup is already a hundred short videos; they are just glued together and in the wrong shape. The obstacle is not editing skill, it is inventory — knowing which product is at 14:22 and what was said about it, a hundred times over.

1. Go to Vizard Agent and upload the full roundup.
2. Say whether you want one short per product or themed sets.
3. Say the length, the shape and whether you want captions.

What do you need before you start?

The long video and any product information you already have. Vizard Agent builds its own map of the video, and supporting material — a product list, links, prices — makes the captions accurate rather than approximate, since a transcript can only carry what was actually said aloud.

• The full roundup, uncut. Vizard Agent maps the whole thing before cutting anything.
• The best copy you have. Resolution matters here more than usual, and the reason is in the limits section below.
• Your product list, if you have one. Names, links and prices that were never spoken aloud can only come from you.
• How to split it. One short per product, or grouped by category, or just the ten best.
• The destination. Vertical for feeds, and say whether captions should be burned in.

What do you type into Vizard Agent?

Say how you want it divided. That is the decision Vizard Agent cannot make for you — a hundred separate shorts and ten themed compilations are both reasonable readings of the same footage, and they are entirely different pieces of work.

Prompt

Variants worth knowing:

• Themed sets. "Group them into five shorts by category" rather than one per item.
• Just the best ones. "The ten most interesting products only." Vizard Agent has the whole map, so it can select.
• Keep my narration. The default. Say so explicitly if the voice is part of your channel's identity.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real thirty-three minute roundup. The middle steps are the ones that make the rest possible: it turns a long video into a written index before it makes a single cut.

1. Takes a first look at the source video and the project files.
2. Checks the analysis tools and samples frames from the source.
3. Grabs sample frames and views them to see what the footage looks like.
4. Maps every shot and transcribes the narration across the full thirty-three minutes.
5. Reads the complete shot log and transcript to understand all hundred products.
6. Reads the narration transcript again specifically for product names and details.
7. Saves the full transcript to a file so later steps work from it rather than re-deriving it.
8. Searches for a higher-quality copy of the original video, because the source it was given was low resolution.
9. Lists the available asset tools and their prices before committing to an approach.

Step 8 is worth dwelling on. Vizard Agent noticed the source was not good enough for what was being asked and went looking for a better one rather than quietly delivering soft footage.

What does the result look like?

Output resolution follows the best source available, which in this category is often the binding constraint. In the run behind this page the uploaded copy was low resolution, and although Vizard Agent went looking for a better version, nothing it produces can carry more detail than the file it was given.

Timing scales with length. This category was not separately measured, and comparable work runs a median of 28 to 38 minutes end to end; a thirty-three minute source with a hundred products sits well beyond that, because the mapping and transcription happen before any cutting starts.

When does this not work well?

Everything here depends on the source file, and roundup videos are frequently re-uploads, screen recordings or compressed downloads rather than the original export. Vizard Agent works with what it is given, will go looking for something better, and cannot invent detail that was never in the file to begin with.

• A low-resolution source stays low resolution. Upload the best copy you have, not a re-download. This is the single biggest determinant of whether the shorts are usable.
• Products that are only named, never shown. If the narration mentions something the camera never frames, there is no shot to cut.
• Prices and links go stale. Vizard Agent puts on screen what you gave it. A price from a month ago is your claim to the viewer, not the video's.
• A hundred shorts is a hundred jobs. Ask for a manageable batch first, check the shape, then continue.
• Affiliate disclosure is yours. If these are affiliate products, the disclosure requirement travels with the clips, and it will not appear unless you ask for it.

How do you fix a result that came back wrong?

Name the product or the timestamp. Vizard Agent keeps the shot map and the saved transcript, so a re-cut works from an index it has already built — which matters enormously here, because rebuilding that index for a thirty-three minute video is the expensive half of the job.

• "The clip for product 40 starts too late." Re-cut from the same map.
• "Use the product names from my list, not the transcript." Captions re-rendered without re-analysing.
• "Make these ten vertical instead." A re-frame of clips that already exist.

How does Vizard Agent compare to doing it yourself?

By hand this is a spreadsheet and a scrub bar: watch the whole video, log each product with its timecode, then cut a hundred clips one at a time. The cutting is trivial. The logging is the day, and it is the reason most roundup channels post one long video and nothing else.

Common questions

Can it really handle a thirty-minute source? Yes. Vizard Agent maps and transcribes the whole thing first, which is what makes the rest possible.

Will it keep my narration? Yes, by default. Say if you would rather have new narration or none.

Can it put the product name on screen? Yes, and it is worth giving Vizard Agent your written list so the names and spellings are yours rather than the transcript's guess.

Can it work from a YouTube link instead of an upload? Yes, and Vizard Agent will fetch the best version available at that link, which is often better than a copy that has been downloaded and re-uploaded.

Should I upload the highest quality version I have? Always. Output detail is capped by the source, and this is the category where that cap bites hardest.