One shoot, several finished pieces

The economics

The expensive parts of a video are the shoot and the structural edit. Once those exist, a second and third deliverable cost a fraction of the first — if the first was built with the others in mind.

A typical set from one interview shoot:

• The long piece. Eight to twelve minutes, horizontal, for a website or YouTube.
• Two or three vertical shorts, 30 to 60 seconds each.
• A square or vertical cut of 90 seconds for a feed post.
• A 15-second teaser that links to the long piece.
• Stills, pulled as frames, for thumbnails and posts.

That is five deliverables from one edit. Quoting for one and delivering five is how people lose money; quoting for the set is how the work becomes worth doing.

Short is not a trim of long

The mistake is taking the ten-minute film, cutting forty seconds out of the middle, and posting it. It does not work, because the two forms have different requirements.

The long piece can build. It can introduce people, establish context, hold a question open for two minutes.

The short piece has to be complete in itself. It opens on the thing, delivers one idea, and ends. There is no context to draw on because the viewer has none.

So a short is a separate piece built from the same material with its own spine — usually one moment or one claim from the long piece, given a beginning and an end of its own.

Building the short version

1. Find the single strongest 20 to 40 seconds in the long piece. Usually one answer, one demonstration, one moment.
2. Ask what a stranger needs to know to understand it. Usually one sentence — a name, a place, a number. That becomes a caption or a super, not a spoken introduction.
3. Open on the moment, not on the setup.
4. End on the end of the moment. No sign-off, no "watch the full video", no logo — or if there must be a call to action, one line of text over the last two seconds, not a five-second card.
5. Captions, burnt in, in the middle third of the frame.

Reframing 16:9 to 9:16

The mechanics, from the resolution lesson: duplicate the timeline, change its resolution to 1080×1920, and reframe each shot.

Three approaches, in increasing order of effort:

• Automatic subject tracking. Resolve's Smart Reframe, Premiere's Auto Reframe, CapCut's equivalent. Fast, and the output is a first pass — it locks onto the wrong subject in any shot with two people or heavy movement. Check every shot.
• Static reframe per shot. Decide one position for each shot and leave it. Usually better than a tracker that wanders.
• Hand-keyframed pans. For shots where the subject genuinely crosses the frame. Slow, and only worth it on the few shots that need it.

What cannot be reframed: a two-shot where both people matter. Your options are to crop to one of them, to stack the frame with picture above and captions below, or to use a different shot.

If you also shoot, the cheap fix is to compose with both crops in mind — keep the subject in the central square and avoid putting anything essential at the frame edges. It costs nothing and it makes the vertical version automatic.

Where the assets come from

The parked timeline from the previous lesson is where the shorts live. Material cut from the long piece for being tangential is frequently perfect as a standalone short, because tangential in a structure means self-contained on its own.

Pull thumbnails as frame exports from the graded master, not from the camera original — you want the colour to match the film.

Order of operations

Make the long piece first, to picture lock, then derive. Deriving from an unlocked cut means redoing the derivatives when the master changes.

The exception is when the short is the actual deliverable and the long version is secondary, which is increasingly common in commercial work. In that case build the short first with its own spine, and treat the long piece as the derivative.

A note on the platforms

Specifications for length, aspect and file size differ per platform and change several times a year. Do not learn them; look them up at delivery time from the platform's own help pages, and keep a note of what you used per client in the project's docs folder.

What has been stable for several years: 1080×1920 for vertical, H.264 in an MP4, captions burnt in, and the safe-area rule about the bottom fifth.

Today

Take something you have published and build one 30-second vertical cut from it with its own beginning and end. Note how little of the original structure survives.