How to edit short-form videos: the complete workflow

Every short cut from a longer video goes through the same eight steps. Do them in this order: captions go on after the cut, or every trim throws them out of sync, and the sound is mixed last, once nothing else moves.

1. Pick the moment: one idea that makes sense to someone who's never seen the long video.
2. Cut it tight: start on the hook, end on the payoff, and take out the pauses and filler in between.
3. Frame it 9:16: crop to the speaker, or stack two shots, at 1080 × 1920.
4. Caption every word: a few words at a time, big and high contrast.
5. Add hook text: a short line on screen in the first seconds that says why to stay.
6. Pace it: punch-ins, B-roll and sound effects where the picture would otherwise sit still.
7. Mix the sound: the voice clear and even, any music well under it.
8. Export: an MP4 at the right size, frame rate and bitrate.

By hand, the slow steps are the first four: finding the moment and captioning it take the longest. The steps that make a short feel edited are the last four. The checklist at the end has the time each one takes.

A short is one idea: a story with a payoff, a strong opinion, a surprising answer, a mistake and what it taught. It has to work for someone who's never heard of the speaker. Let the moment set the length: most clips that work land between 20 and 60 seconds (more in how long a short video should be).

• Read, don't watch. A transcript lets you skim an hour in minutes, and the strong lines stand out on the page. The free video to text converter makes one in your browser.
• Work in pieces. Break a long recording into parts with the video splitter and go through them one at a time, so you're never scrubbing a three-hour file.
• Start on the hook. Open on the line that makes someone stay: a claim, a question, a number, the result first. Cut the greeting, the “so, yeah” and the setup before it. Short video hooks has templates for each kind.
• End on the payoff. The punchline, the answer, the reveal. Cut before the laugh trails off into the next topic.

The first seconds are where the apps measure you. YouTube's Shorts analytics report “Stayed to watch”: the share of times viewers watched past the first seconds of a Short. TikTok's Creator Academy says the same thing from the other side: put an exciting hook right at the beginning, so viewers are curious enough to keep watching.

A jump cut joins two pieces of the same shot with the part between them taken out. In short-form editing it's the main edit: every pause, false start and “um” you remove leaves one, and a quick run of them is what makes a talking head feel fast.

• Pauses. Take out silences longer than about half a second. The free remove silence from video tool finds the quiet parts for you, in your browser; how to remove silence from a video explains its settings.
• Filler words. “Um”, “uh” and “you know” are sound, not silence, so silence removal leaves them in. Editors that work from a transcript find them: in Adobe Premiere (Premiere Pro), the Text-Based Editing workspace's Transcript panel has a filter for filler words and pauses, with Delete all; CapCut's Auto captions can highlight filler words in the transcript for you to delete.
• False starts and repeats. When a sentence is said twice, keep the better take and cut the other.
• Breaths. Don't cut every gap to nothing. Leave a tenth of a second or so either side of each cut, or the speech runs together and sounds robotic.
• Whole sentences. Cut between sentences or clear phrases, never mid-word. The video trimmer takes a clip down to its start and end without losing quality.

TikTok, Reels and Shorts fill a phone screen at 9:16: 1080 × 1920 pixels. A 16:9 recording has to lose most of its width, and choosing what stays is the framing.

• Crop on the speaker. With one person talking, crop to their face and follow them as they move. With two, switch to whoever is speaking.
• Stack. When both people matter at once, or for a streamer's webcam over the game, put one crop above the other. The split screen video maker does it in your browser.
• Skip the bars. Fitting the whole wide frame with blank space above and below leaves faces small. See video sizes for TikTok, Reels and Shorts for what each option keeps.
• Keep text in the safe zone. The apps lay their buttons, username and caption over the edges. Keep captions and hook text inside the 820 × 1260 px area that's clear on all three: 240 px from the top, 420 from the bottom and 200 from the right. Check a frame against each app's layout with the safe zone checker.

The free vertical video converter crops or fits a video to 1080 × 1920 in your browser. In Premiere, Sequence > Auto Reframe Sequence makes a copy of a sequence at the aspect ratio you pick and follows the action; its Slower Motion setting is meant for talking-head interviews. DaVinci Resolve Studio, the paid version, includes Smart Reframe.

Plenty of people scroll with the sound off, so the captions carry the clip. What works on a phone:

• Word by word, a few at a time. Three words on screen is plenty, with the word being said in an accent color.
• Big and high contrast. A heavy sans-serif, white or bright, with a dark outline or shadow so it reads over anything.
• In the lower middle, below the face and above the apps' bottom margin.
• Every word right. Automatic captions get names, brands and numbers wrong most often. Check each one.

The free auto caption generator transcribes a clip in your browser into timed captions you can fix, and add captions to video burns them in word by word. In Premiere, choose Create captions from transcript in the Text panel's Captions tab, and set Layout to Single Word for one word at a time. In CapCut, it's Captions > Auto Captions, the language, then Generate. More in how to caption short videos.

Hook text is a short line on screen in the first seconds that says what the viewer gets for staying. TikTok's Creator Academy suggests text on screen as a supplement to the sound and picture: a brief title or a question that makes people curious. Keep it to a handful of words, in the top third and inside the safe zone, and don't just repeat what's being said.

Once the clip is tight, give the eye something new every few seconds, most of all where the picture would otherwise sit still. These are the usual pattern interrupts:

• Punch-ins. Push in on the line that matters, and pull back for the next.
• B-roll. A few seconds of what's being described: the product, the place, the screenshot, the chart.
• Text and graphics. The number on screen as it's said, an arrow at what they're pointing at.
• Sound effects. A whoosh on a cut, a pop when text appears. Use them sparingly: if everything has a sound, nothing stands out.

Clyper doesn't add B-roll, zooms, music or sound effects. It finds, cuts, frames and captions the clip; these finishing touches are yours to add in your editor afterwards.

Sound is the step most often skipped, and the one people notice when it's wrong:

• The voice first. Clear and even from start to finish, with no clipping on the loud words. Fix a quiet speaker before you add anything else.
• Music well under the speech. Low enough that every word is easy to follow, and lower still while someone talks. In Premiere, the Essential Sound panel can duck music under dialogue for you.
• Licensed sound only. Use music you have the rights to, or the apps' own libraries: TikTok's web editor lets you add licensed sounds and check a video's music before you post.
• Even from clip to clip. A feed of clips should play at about the same loudness. Clyper evens out every clip's sound to −14 LUFS; in an editor, a loudness meter shows where yours sit.

Post the best copy you have: don't compress a clip before uploading it, since each app compresses it again. On Instagram, turn on Upload at highest quality (Profile, then Menu, then Media quality). For every app's sizes and limits in one place, see video sizes for TikTok, Reels and Shorts, or check a file with the video size checker.

Clyper's clips come out as 1080 × 1920 MP4s with H.264 video and AAC sound, at the source video's frame rate up to 60 fps, with no watermark.

How long it takes. Here are rough times for 5 clips from an hour-long video you haven't watched, for an editor who knows their app. They're our estimates, not measurements, so swap in your own:

“With Clyper” is the time you spend checking its work, not counting the wait while it processes the video. Clyper takes the first four steps: it reads the transcript to find the best moments and scores each one's hook, cuts each clip on whole sentences, frames it 9:16 on whoever's talking (switching between speakers, and putting a streamer's webcam over the game), and captions every word in 50+ languages, in 5 styles. In the clip editor you can trim or extend either end and fix any caption word. What's left is the polish that makes a clip yours: hook text, B-roll, effects and music.

If you edit for clients, those saved hours are your margin. How much to charge for clipping works through what to quote. Starter is $20 a month for 150 credits. 1 credit = 1 minute of source video.