# Short form video editing for the clip on your phone

The order to edit in, what to cut, where the hook and captions go, how to export for each platform, and how to hand the mechanical half to AI while you keep the decisions.

Source page: https://postmosis.com/vault/short-form-video-editing. By Leo at Postmosis. Last checked 2026-10-29. Adapted from Keepy Labs: [Short-Form Video Editing with Astra and After Effects](https://keepylabs.com/chatgpt-codex-guide/short-form-video-editing).

In short: Edit a short in this order: cut everything before your first word and every pause, retake and filler word, open on the hook, add captions, cover abstract lines with a cutaway, then export upright at 9:16 [1]. Keep it under 3 minutes if you want Instagram to show it to new people [2]. AI can do the cutting and captions, while the hook and the final watch stay with you [3].

## Edit in this order

Say you filmed a quick answer to a client question between meetings. Whatever app you use, it goes together in the same order. Work out of order and you end up captioning lines you later cut.

1. Cut the start, the pauses, the retakes and the filler.
2. Move your strongest line to the front and put it on screen.
3. Change the screen when the idea changes: a cutaway, a comparison, a number.
4. Add captions.
5. Fix the sound, and add music only if it helps.
6. Export upright and watch it once on your phone.

Example: A career coach films a 2-minute answer to 'how do I ask for a raise?' between sessions. She cuts the 4 seconds of setup at the start and 11 pauses, which brings it to 70 seconds. Her best line, 'Ask for the number, not for a review', was at 0:48, so it moves to the front and goes on screen. She covers 'look at what the market pays' with a screenshot of a salary survey, adds captions in one style, turns the background hum down and exports at 9:16. It goes up after one watch on her phone with the sound off.

## Film so the edit is easy

The easiest edits start at the camera.

- Film upright at eye level, with the light in front of you. The [talking head video guide](https://postmosis.com/guides/talking-head-video) has the full setup.
- Stay quiet for a moment before your first word and after your last, so the cut points are clean.
- If you fluff a line, stop, pause, and say the whole sentence again, so you can cut to the clean version in one go.
- Record your hook twice on its own at the end, while you're warmed up.
- Before you stop, film a few seconds of B-roll: your hands on the work, your desk, the place you're talking about.

## Cut the dead air

Most of the edit is taking things out: the breath before you start, the false start, the um, the pause while you remember the next point. Each cut that jumps you forward mid-sentence is a [jump cut](https://postmosis.com/glossary/jump-cut), and on a talking video nobody minds them.

- Start on your first word. When you post a reel on Instagram, the Trim slider cuts the beginning and end [2].
- Cut every pause longer than a breath. Leave one short beat after a line that should land.
- Keep the last take of any line you said twice. When you film, pause before you repeat a line so the cut is easy to find.
- Cut sentences that repeat the one before. If you made a point twice, keep the shorter version.

If your editor shows a transcript, cut from that. Repeated sentences are easier to spot in text than on a timeline. The [talking head video guide](https://postmosis.com/guides/talking-head-video) covers filming so there's less to cut in the first place.

## Open on the hook

The first thing on screen should be the reason to keep watching. Start on your strongest line, even if you said it halfway through, and put it in large text near the top of the frame for anyone watching on mute.

If you didn't film a strong opening, write one and record it on its own; a five-second clip of you saying it is enough. The [hook generator](https://postmosis.com/tools/hook-generator) gives you openings to try, and [video hooks](https://postmosis.com/vault/video-hooks) breaks down the hook types that hold people.

Example: A family lawyer's raw clip starts: 'So, um, people ask me about prenups a lot.' The edit opens instead on a line from later in the clip, 'The prenup question nobody asks until it's too late', with that sentence in large text at the top.

## Make the screen follow the idea

A talking clip loses people when you move to a new point and the picture stays the same. The edit's job is to change the screen when the idea changes: key words get emphasis, abstract claims become pictures, comparisons sit side by side, and a sound marks the big transitions [3].

For each beat, pick one of these [3]:

- Captions with the key word picked out in colour or size.
- A cutaway, or [B-roll](https://postmosis.com/glossary/b-roll): a few seconds of the thing you're describing.
- A comparison: before and after, or two options side by side.
- A number on screen when you say one.
- A simple graphic, such as a step card or a progress bar.
- A punch-in: a quick zoom on your face for the line that matters.

Skip effects that look busy and don't help anyone understand you [3]. A spinning transition gives people something to watch other than you, right when you're making your point.

Example: What that looks like in your work: a consultant comparing two ways to price a project shows them side by side. An agent describing the kitchen cuts to the kitchen. A coach who says 'three sessions in, most clients want to quit' puts '3 sessions' on screen as she says it.

## Captions that read on a phone

Captions carry the video for anyone watching on mute. Instagram can make them from your speech [4], and most editors can too. Check every word, especially names and numbers.

- Keep each caption short, two to five words, so the eye keeps up.
- Make them big enough to read with the phone at arm's length.
- Keep them clear of the bottom of the frame and the right edge, where the post caption, the buttons and your name sit.
- Stick to one caption style for the whole video, because a style that changes halfway through looks like a mistake.

Review the edit on your phone, not your laptop: check caption size, pacing, safe zones and whether each visual lands on the right word [3].

## Fix the sound

People will sit through a plain background long before they'll sit through bad audio. If your voice is quiet or echoey, fix that before anything else; most editors have a noise or voice clean-up setting.

- Keep music well under your voice, or leave it out of a talking video.
- Use a short sound on the big transitions, such as a soft whoosh into a new point [3].
- Use music you have the rights to. On YouTube, a Short over one minute with any active copyright claim is blocked worldwide [5].

## Export for each platform

Export upright at 9:16, the tallest shape Instagram takes for reels [1]. The same vertical file also counts as a YouTube Short [5].

Short-form limits by platform:

| | Shape | Length |
| --- | --- | --- |
| Instagram Reels | 1.91:1 to 9:16, at least 720 pixels and 30 frames a second | Up to 20 minutes; reels over 3 minutes won't be recommended to new audiences [1][2] |
| YouTube Shorts | Square or vertical | Up to 3 minutes [5] |
| TikTok | Your upright master | Everyone can post 3-minute videos, and some accounts can post longer [6] |

For every network's limits in one place, see [social media video length](https://postmosis.com/guides/social-media-video-length), and [aspect ratio](https://postmosis.com/glossary/aspect-ratio) if the shapes are new to you.

Postmosis edits the clip you filmed on your phone in the browser: it cuts the pauses, retakes and filler words, adds captions in 6 styles, keeps your face in frame, drops in B-roll and exports 9:16, 4:5, 1:1 and 16:9 in one run. [See the video editor](https://postmosis.com/features/video-editor)

## Best editors for your week

You don't need the most powerful editor. You need the one that speeds up your slowest step, and for anyone who talks to camera that's usually cutting pauses and captioning. Start from how your week of posting actually looks.

Which editor fits your week:

| | Try | Why it fits |
| --- | --- | --- |
| You film one clean take and post it straight to Instagram | Instagram's own editor | It trims the start and end and makes captions from your speech [2][4] |
| You edit on your phone between client meetings | Edits, Instagram's video app | Captions, a voiceover read from an on-screen script, music and sound effects, then export straight to Instagram [7] |
| You ramble, and the ums are the slow part | Descript | It finds the ums and uhs in your transcript and lets you remove or keep each one, in English transcripts only [8] |
| Your clients speak another language, or several | CapCut | It builds captions from speech recognition and can translate them [9] |
| Your carousels and graphics already live in Canva | Canva's video editor | Hook templates for social video and captions you generate in the same place [10] |
| You post the same kind of video every week with your brand graphics | After Effects with an AI agent | It builds the layers from a plan you approve, and the tenth video reuses what the first one decided [11][3] |

Whichever you pick, the order of the edit stays the same, and the hook, the argument and the final watch stay with you [3]. Try one editor on three of your own clips before you switch to another.

## Use the same shape every week

People start to recognise you when your shorts look alike. Pick one structure and keep it: the hook on screen, the point, one example or piece of proof, then one next step. The [video script template](https://postmosis.com/templates/video-script) gives you that shape before you film.

Keep the same caption style, the same text position and the same end card on every video. The edit gets quicker too, because you reuse the settings instead of choosing them each time.

## When AI does the editing

By the tenth tip video you've placed the same title card ten times. That's the point where AI earns its place. It's good at the mechanical half: transcripts, cutting silences and filler, captions and a rough first cut. The decisions stay with you: the hook, the argument, the proof, your taste, your brand's standards and the final approval [3].

There are two levels. A phone or browser editor with AI features does the cutting and captions, one clip at a time. The heavier route has an AI agent plan the edit and build branded graphics in After Effects, which pays off once you're making the same kind of video every week.

The agent route works because Codex, OpenAI's coding agent, can inspect files, make edits and run the tools on your computer [12][13]. After Effects can be driven by scripts written in a version of JavaScript, made for automating repeated tasks [11]. So the agent can build the layers you'd otherwise place by hand, from a plan you approve.

## The agent route in four steps

The agent route is four steps, done in order for every video [3]:

1. Prepare one project folder with everything the agent needs, listed below [3].
2. Map the visual beats: a plan, before any editing, that links each spoken beat to a visual [3].
3. Render a 15-second preview and approve the look on that before the full export [3].
4. Give timestamped feedback, then let the agent revise [3].

For step one, the folder holds:

- The raw video and its transcript.
- Your brand colours, fonts, logo and layout rules [3].
- Two or three videos whose look you like [3].
- Any After Effects templates and parts you reuse.
- Footage, pictures, icons and sound effects you already have.

Keep it all in one project so each approved decision improves the next edit [3].

Step two, ask for a plan before any editing. It should link each spoken beat to a visual, say why it helps, and flag where the idea moves but the screen doesn't [3].

Step two prompt: map the visual beats:

```
Look at this short video and its transcript before you edit anything.

Make a timeline table with: timestamp, spoken cue, proposed visual, and the reason it helps. For each important beat, pick one: captions, a cutaway, a comparison, a number, a motion graphic, or one of my existing After Effects components.

Flag the three weakest visual moments, where the idea moves on and the screen doesn't.

Use my brand guide and saved assets first. Don't edit yet. Show me the plan to approve.
```

Step three, approve the look on a short section before the full render [3].

Step three prompt: a short preview first:

```
Use the approved visual plan to make a 15-second After Effects preview.

Follow the brand guide and the existing templates. Keep me readable on screen. Every visual change has to support the spoken idea; leave out decoration that doesn't help someone understand it.

Show me a few representative frames and the preview before you render the full video.
```

Step four, watch the preview on your phone and give notes with timestamps. Concrete notes get better results than asking it to make it more engaging [3].

Step four example: timestamped notes:

```
00:04-00:08: Replace the generic cards with the photos in the assets folder.
00:09-00:12: Keep the captions, but slow the transition.
00:13-00:15: Show the comparison side by side instead of one after the other.
Leave everything else as it is.
```

After the four steps, package the parts you repeat most, so the next video reuses them:

- A hook title and a progress bar, so every video opens the same way.
- Number counters, step cards and before-and-after comparisons.
- Phone and browser frames for when you show a screen.
- Your caption styles and safe zones, saved as presets you load in one click.
- End cards, transitions and sound cues.

The first video takes decisions, and the tenth should reuse them [3]. After each approval, have the agent separate one-off fixes from preferences you'll want every time, and save only the ones worth keeping to a shared editing guide [3].

Prompt: save the rules:

```
Review the feedback I approved on this edit.

Split it into one-off changes and preferences I'll want on every video. Update the shared editing guide with only the rules worth keeping for pacing, type, captions, motion, audio and visual patterns. Show me exactly what you plan to save before you change any project files.
```

Checklist for every AI-assisted edit:

```
[ ] Raw clip, transcript, brand guide and references in the project folder
[ ] Timestamped visual plan approved
[ ] 15-second preview watched on my phone
[ ] Notes given with timestamps
[ ] Lessons worth keeping saved to the editing guide, and one reusable part packaged
[ ] Full export approved
```

## When editing can't save a clip

Some clips are quicker to film again than to fix. Refilm when:

- You can't say the point in one sentence. No cut makes a muddled answer clear, so write the sentence and film that.
- The sound is echoey or the wind took it. A clean-up setting helps a little; a quieter room helps more.
- It's well over 3 minutes and you want new people to see it. Instagram won't recommend reels over 3 minutes to new audiences [2], so split it into two posts or refilm it shorter.
- There's no hook anywhere in it. Record one line on its own and open on that.
- You read it from a script and it shows. Film it again in your own words, with your eyes on the lens.

## Common editing mistakes

- An intro before the point: a logo, a 'hey guys', a slow zoom. Start on the hook.
- Captions hidden under the buttons or the post caption. Check the frame on your phone before you post.
- Music louder than your voice. Turn it down until every word is easy to follow with your eyes shut.
- A landscape export for a vertical feed, which leaves a thin strip of video in a mostly empty screen.
- Cutting straight from a strong line into the next point, so it never gets a beat to land.
- Posting without one full watch on a phone, once with the sound on and once with it off.

## Cut shorts from a longer recording

A webinar, a podcast or a recorded client Q&A holds several shorts. Look for moments that make sense with no context: a complete answer, a story with an ending, an opinion you'd defend. Start each clip on the first word of the thought rather than on the question before it, and give each one its own hook on screen.

Then edit each clip like any other short: cut, hook, captions, export. [Repurpose content](https://postmosis.com/vault/repurpose-content) covers turning one recording into a week of posts.

Postmosis finds clips inside a recording of 45 seconds or more and opens 3 to 5 of them as their own projects, ready to caption and export. [See the video editor](https://postmosis.com/features/video-editor)

## Common questions

### What is short form video editing?

Turning raw footage into a short vertical video for Reels, TikTok or YouTube Shorts: cutting the pauses and retakes, opening on the hook, adding captions and cutaways, and exporting upright. On YouTube, vertical videos up to three minutes count as Shorts [5].

### What's the best AI video editor for Reels?

One that cuts pauses and filler from a transcript, captions your speech and exports at 9:16 [1], so you aren't rebuilding the edit by hand. Whichever you pick, you still choose the hook and watch the result before it goes up [3].

### How long should a short-form video be?

As long as the point needs. Instagram won't recommend reels over 3 minutes to new audiences [2], and YouTube counts vertical videos up to three minutes as Shorts [5]. Time your script before you film with the [script length calculator](https://postmosis.com/tools/speech-time-calculator).

### Can I edit a short-form video on my phone?

Yes. Instagram lets you trim the start and end of a reel [2], and it can caption your speech automatically [4]. For cutting pauses mid-clip and adding cutaways, a separate editor is quicker.

### Should I add music to a talking video?

Only quietly, under your voice, and only music you have the rights to. YouTube blocks Shorts over one minute that carry an active copyright claim [5]. Plenty of talking videos work fine with no music at all.

### Can AI edit a whole video for me?

It can cut, caption and build graphics from a plan you approve. Keep the hook, the argument and the final approval for yourself [3], and watch a short preview on your phone before the full export.

### How do I add captions to a Reel?

Instagram uses speech recognition to make closed captions for reels automatically [4]. Check them for misheard names and numbers, or add your own captions in the edit so they look the same on every platform.

## Sources

1. [Reel size and aspect ratios on Instagram, Instagram Help Centre](https://help.instagram.com/1038071743007909), read 2026-10-29
2. [Record a reel on Instagram, Instagram Help Centre](https://help.instagram.com/225190788256708), read 2026-10-29
3. [Short-form video editing with Astra and After Effects, Keepy Labs](https://keepylabs.com/chatgpt-codex-guide/short-form-video-editing), read 2026-10-29
4. [Manage closed captions for your reels on Instagram, Instagram Help Centre](https://help.instagram.com/7487270478066359), read 2026-10-29
5. [Understand three-minute YouTube Shorts, YouTube Help](https://support.google.com/youtube/answer/15424877?hl=en), read 2026-10-29
6. [Content Posting API media transfer guide, TikTok for Developers](https://developers.tiktok.com/doc/content-posting-api-media-transfer-guide), read 2026-10-29
7. [Edit your video on Edits, Instagram Help Centre](https://help.instagram.com/583491847610362/), read 2026-10-29
8. [Filler words, Descript Help Center](https://help.descript.com/hc/en-us/articles/10164806394509-Filler-words), read 2026-10-29
9. [Smart auto caption generator, CapCut](https://www.capcut.com/tools/auto-caption-generator), read 2026-10-29
10. [Free online video editor, Canva](https://www.canva.com/video-editor/), read 2026-10-29
11. [Scripts in After Effects, Adobe Help Center](https://helpx.adobe.com/after-effects/using/scripts.html), read 2026-10-29
12. [Codex CLI, OpenAI Developers](https://developers.openai.com/codex/cli), read 2026-10-29
13. [Using Codex with your ChatGPT plan, OpenAI Help Center](https://help.openai.com/en/articles/11369540-codex-in-chatgpt), read 2026-10-29