How to make shorts from several YouTube links at once

Last updated 2026-08-24 · 7 min read

What is the short version?

Cutting shorts from six episodes is not six separate jobs. The hooks compete with each other across the whole set, and the crop has to be solved per episode because a multi-panel show rarely puts the speaker in the same place twice.

1. Go to Vizard Agent and paste all the links.
2. Say you want vertical shorts with captions.
3. Say roughly how many, or let Vizard Agent decide from what it finds.

What do you need before you start?

The links. Vizard Agent downloads, transcribes and searches all of them, so a list is the whole brief. Saying what makes a good hook for your audience helps, since the strongest moment in a conversation is not always the most quotable one.

• The links. All of them in one message.
• How many shorts. Or leave it to what the material supports.
• What counts as a hook. A claim, a disagreement, a surprising number.
• The shape. Vertical with captions, almost always.
• Any episode to prioritise. If one matters more than the rest.

What do you type into Vizard Agent?

List the links and stop there. Vizard Agent handles the whole batch as a single job rather than as six separate ones, which is exactly what lets it compare candidate moments across every episode instead of picking the best one from each in isolation.

Prompt

Variants worth knowing:

• A fixed number per episode. If you need even coverage rather than the best overall.
• Titles per clip. Vizard Agent reads the source titles and can carry them through.
• One loudness target. So the batch plays evenly when posted as a series.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real batch. The part that separates a batch job from six small ones is the middle: the panel geometry of every source was measured numerically before a single crop was chosen.

1. Downloads all six source videos at once and probes them.
2. Samples frames from each and reads the source titles.
3. Starts transcription of all six in the background and checks the results as they land.
4. Hunts hooks across all six transcripts together rather than one at a time.
5. Reads the exact wording around every candidate moment in each episode.
6. Measures the picture area in the frames and builds a comparison image of a host shot against a show shot.
7. Measures the panel geometry in each source and maps the positions numerically, inspecting the layouts that differ.
8. Generates styled captions for all the clips and renders thirteen vertical shorts.
9. Reviews the renders, re-renders with corrected framing, levels every clip to a single loudness target, adds a limiter, then fixes two clips whose in and out points were wrong and re-cuts one splice after watching it back.

Step 7 is what stops a batch looking careless. Each episode of a panel show lays the frame out slightly differently, and a single crop setting applied across all six will centre on the guest in one and on empty desk in another.

What does the result look like?

From the run this page is written from, probed on one delivered short: 1080x1920, H.264, 30fps, 65.1 seconds, AAC audio. Vertical, just over a minute, one of thirteen clips cut from six source episodes, each framed to its own episode's layout and levelled to a shared loudness target.

Thirteen shorts from six episodes is a little over two per video, which is the honest yield. Vizard Agent cut what genuinely stands alone rather than filling a quota, and levelling them all to one target is what makes them work as a series.

When does this not work well?

A batch inherits every problem that every one of its sources has, and the yield depends on the quality of the conversations rather than on the number of links you paste. Vizard Agent works through all of them properly and cannot manufacture hooks that were never said.

• You need the rights to all of them. A list of links is not permission.
• The yield varies per episode. Some conversations produce three clips and some produce none.
• Panel layouts crop badly. A wide three-person shot does not fit a vertical frame.
• Burnt-in graphics come along. Lower thirds and show branding end up in the crop.
• A long batch takes real time. Six transcriptions and thirteen renders is not instant.

How do you fix a result that came back wrong?

Name the clip that is wrong. Vizard Agent keeps all six transcripts, every candidate moment it found, the panel measurements and each render, so re-cutting one short or reframing another is a targeted re-render rather than a restart of the whole batch.

• "Clip four starts mid-sentence." Re-cut from the word timings, as this run did twice.
• "The crop is off in the Boris episode." Re-framed against that episode's measured layout.
• "They play at different volumes." Re-levelled to a single target with a limiter.

How does Vizard Agent compare to doing it yourself?

By hand this means downloading six videos, watching each one for something worth clipping, and setting a crop that works for the first episode and fails for the rest. The volume differences are the part people miss, because each clip sounds fine on its own and the series jumps around when posted.

Common questions

How many links can Vizard Agent take at once? Six on this run. It downloads and transcribes them in parallel.

How many shorts will I get? A little over two per video, typically. Vizard Agent cuts what stands alone rather than filling a quota.

Will they be framed correctly? Yes. Vizard Agent measures each episode's panel layout rather than applying one crop to the batch.

Do they get captions? Yes, generated per clip from that clip's own word timings.

Why level them all to one target? Because shorts from different episodes are usually posted as a series, and a viewer who adjusts their volume for one clip should not have to adjust it again for the next. Vizard Agent sets a single target across the batch and adds a limiter so nothing clips.

Can Vizard Agent use the source titles? Yes. It reads them and can carry them through to the clips.

What if one episode has nothing worth clipping? Vizard Agent will say so rather than cutting something weak to make the numbers even.

Does Vizard Agent check the batch? Yes. It reviews the renders, has them watched for sync and cut quality, and re-cut one splice on this run.