Producing vertical clips from conference footage shot for a wide stage

The brief said vertical clips from conference footage, and the footage exists: three days, two stages, a house crew with a wide shot that holds the full set so the sponsor logos stay in frame. The editor crops the first session to 9:16 and sends a test. It is a podium, a lectern, half a screen, and a person who reads at phone size as a smudge in a jacket.

Nobody did anything wrong. The camera was framed for the job it was given, which was to record a session for on demand replay with the branding visible. Vertical output was decided afterwards, by somebody who was not in the production meeting, and the geometry does not care.

What a wide master does to a 9:16 crop

Work it in pixels and the argument stops being aesthetic.

A 16:9 master at 1920 by 1080 has a full height 9:16 region 1080 times 9 divided by 16 pixels wide, which is 607.5. Delivering that at 1080 by 1920 means enlarging by 1080 over 607.5, a factor of 1.78. Everything in the frame, including the compression artefacts, gets 78 per cent bigger.

The same crop from a 3840 by 2160 master is 2160 times 9 divided by 16, which is 1215 pixels wide. Delivering at 1080 by 1920 is a reduction to 0.89 of the source, which is the good direction. You also have headroom: punch in to a region 1200 pixels tall and 675 wide, tight enough for head and shoulders, and you are back to an enlargement of 1.6.

So the first decision is resolution on the master, and it is usually free. Ask the production supplier to record 4K on the wide camera. Storage and data wrangling go up, the day rate mostly does not, and every vertical crop for the next two years starts from a different place.

The second decision is what the wide shot contains. A speaker who occupies a tenth of the frame width in the master occupies a third of the width after a full height crop, centred, with empty stage above their head and below their waist. That is the podium problem, and no resolution fixes it, because the information you want was never captured at a useful size.

Why bother with vertical at all?

Worth checking before spending anything. The Reuters Institute's Digital News Report 2025 found that in the United States the share of people consuming any news video weekly has risen from just over half in 2021 to around three quarters today, and that the majority of that consumption is accessed via third party platforms, at 61 per cent, against 29 per cent via news sites and apps.

Read that as a distribution fact rather than a format fact. If two thirds of the video audience for a news brand is reached somewhere other than the brand's own site, then a media property attached to a trade show that publishes only 16:9 replays on its own domain is choosing to be seen by the smaller share. The format follows from where the audience is.

It does not follow that every session needs a vertical cut. It follows that the ones you want distributed off your own property need to be built for the place they are going.

What does a vertical clip actually cost to produce?

Take a three day show with a clip programme delivering 68 vertical clips.

Route one is reframing from the wide master. A locked off wide shot needs manual reframing per clip, because the speaker walks, and a static crop loses them. Call it 22 minutes per clip of crop, keyframing and checking. At an illustrative 55 per hour, that is 20.17 a clip, or 1,371 across 68.

Route two is a second camera, tight on the speaker, locked off, recording 4K for the full run. A camera package with setup and a share of the operator's time at 1,150 a day for three days is 3,450, plus 350 for media and data handling, so 3,800.

The camera does not only buy pixels. It also removes most of the reframing, because the tight shot is already close to the crop you want. Say that drops the per clip editor time from 22 minutes to 4. The saving is 18 minutes on each of 68 clips, which is 20.4 hours, worth 1,122 at the same rate.

Net cost of the camera is 3,800 minus 1,122, which is 2,678, or 39.38 per delivered clip.

That number is the decision. At 68 clips a year the camera is buying quality at roughly 39 a clip, and reasonable people disagree about whether that is worth it. The crossover where the camera pays for itself on editor time alone is 3,800 divided by 20.17, about 188 clips, which almost no single show reaches and a portfolio running the same crew across six shows reaches easily.

My own view is that the camera is worth it at portfolio level and hard to justify for one show, and that the 4K wide master is worth it everywhere because it costs almost nothing and removes the worst outcome.

One warning about the denominator. Delivered clips means clips that left the edit in a vertical format, and it is tempting to count the 16:9 versions too because the camera helped those as well. Keep them out. The camera is being priced against the vertical programme, and folding in benefits to the horizontal output turns a decision you can check into an argument you cannot. If the tight camera genuinely improves the on demand replay, that is a separate line with its own case.

When reframing is the right answer

There are sessions where the second camera earns nothing. A panel of four seated speakers cannot be framed tight by one locked off camera, and cutting between speakers needs either an operator or four cameras. A demo where the screen is the content wants the screen, and a vertical crop of a screen is usually better served by rebuilding the frame in the edit with the screen capture on top and the speaker in a smaller window below.

Reframing also wins when the clip volume from a session is one. Setting up a camera for a room that produces a single 40 second extract is not a good trade at any day rate.

Automatic reframing, the subject tracking feature in most editing suites, changes the arithmetic but does not remove the decision. It is good at following one person against a plain background and poor at a stage with a moving screen, two people and a moderator's hand entering frame. The honest way to price it is to run it on ten clips from your own footage, count how many needed manual correction, and use that share to adjust the 22 minutes rather than assuming the tool takes it to zero.

The pattern that holds up is to put the tight camera on the stage that produces most of the clips, which is usually the keynote stage, and reframe everything else. Track the clips delivered per stage for one edition and the allocation answers itself the following year.

Delivery specs that decide the crop

Set the output format before the edit rather than after it. YouTube's own documentation for Shorts describes short form videos of up to 3 minutes and a maximum upload resolution of 1080p, which means a 1080 by 1920 delivery is the ceiling and anything above it is discarded on ingest.

That has a practical consequence for the punch in decision above. If the platform tops out at 1080 wide, then a 1215 pixel wide crop from a 4K master is already more than the platform will keep, and the argument for 4K is about the freedom to punch in and stabilise, not about delivering more pixels than the platform accepts.

The safe capture and grading targets are the ones the widest platform in your set will accept. Whether each clip then carries markup so it can surface as a key moment in search is a separate build with its own required fields.

Where this stops

The geometry argument assumes the problem is framing. Often it is lighting. A stage lit for the room, with a bright screen behind the speaker and nothing in front, produces a silhouette that survives a 16:9 replay and dies on a phone in daylight. Cropping tighter makes it worse, because the crop removes the bright screen that was carrying the exposure and leaves an underexposed face filling the frame.

Audio is the other one. A clip that plays without sound in a feed needs captions to work at all, and captions on a 9:16 frame occupy real space that the crop already made scarce. Budget for the caption band in the framing decision, not after it, which means the speaker's head sits higher in the frame than it would in a 16:9 shot.

Neither of these is solved by the second camera, and both of them are cheaper to fix in the production meeting than in the edit.

Pull one recording from your last show, crop a 30 second window to 9:16 in whatever tool is to hand, and view it on a phone at arm's length. If the speaker is unreadable, you have the resolution and framing conversation to have with production before the next show, and you can price it against the clip volume you are already committed to in the wider clip programme. Check at the same time that the release covers a reformatted extract, because a vertical crop with your titling on it is a derivative work, and that is a permission question the content operation should settle once for every show rather than per clip.