AI Video Editing Flows

Editing workflows for Shorts and Reels

This section is about editing footage you already have into vertical video, not about generating video from nothing.

Eight seconds of the 1080x1920 vertical Short for clip tos_a with three-word captions, scaled down for the page
A frame from our own test run: We turned a wide film clip into a vertical Short with free tools and checked the safe zone.

Open the test run

The job and the deliverable

The job: edit footage you already have into vertical video for YouTube Shorts, Instagram Reels and TikTok. This is editing and reframing, not generating video from a prompt.

The deliverable: a 9:16 MP4, 1080x1920, with a hook in the first seconds, captions and any text kept inside the safe zone, steady audio, and a file small enough to upload from a phone.

What you need

  • Source footage with the subject large enough to survive a crop to vertical.
  • A hook line and a single idea for the clip.
  • The safe zone for the platform. Google publishes one for vertical ads; Meta's figures here come from third-party guides; TikTok publishes no fixed pixel zone. See Known problems.
  • A caption method (see AI caption workflows for real footage).
  • A phone to review on.

Recommended workflow

  1. Operator: choose the one idea and write the first line. Everything before the hook goes.
  2. Tool: reframe 16:9 to 9:16. A fixed crop with FFmpeg works for a static subject. For moving subjects, face-tracking tools exist (MediaPipe is a common base; the library lists AI YouTube Shorts Generator and yt-short-clipper, which is Windows only). PySceneDetect can find shot changes so each shot gets its own crop. Operator: check every crop.
  3. Tool: tighten pace by removing dead air, for example with auto-editor. Operator: keep the pauses that carry meaning.
  4. Tool: add captions and text inside the safe zone. For effects that suit short video, see Kinetic typography slam and Whip pan transition.
  5. Tool: normalize loudness with a tool such as ffmpeg-normalize and export with FFmpeg; the command builder at FFmpeg command builder helps write the command.
  6. Operator: watch the export on a phone, with the platform interface visible if you can, and run the checklist.

What we ran ourselves

We ran steps 2 to 5 as one script on three 60-second clips of Tears of Steel: auto-editor to tighten, FFmpeg's scene score to find the shots, a 9:16 crop window chosen for each shot, faster-whisper captions burned in after the crop and loudnorm to -14 LUFS. One cold run and three timed runs per clip, on a laptop CPU. The script, the numbers and the frames we looked at are on our wide-to-vertical test.

  • A whole run took a median of 188.4, 212 and 205.4 seconds per clip.
  • 78 of 78 captions sat inside Google's published vertical safe zone, and 78 of 78 were in the picture.
  • The crop window was 300 px wide in a 1280 px source, so each shot was enlarged 3.6 times. The rule that chose each window (most edge detail) does not know where the subject is; the test page shows where it missed.
  • A person still has to check every crop. We did not test face tracking.

Not run by us: OpusClip, Submagic, the face-tracking clippers in the reviewed picks, PySceneDetect, ffmpeg-normalize, and the vertical tools inside editing apps.

Alternative route

Hosted clippers and caption editors, such as OpusClip and Submagic, do reframing, captions and zooms in a web editor. OpusClip prices and limits are on OpusClip pricing and usage limits. They fit when you produce many clips and want a quick first pass. The open route fits when you want exact control over crops and captions or cannot upload footage to a third party.

Your editing app's own vertical sequence and auto-reframe tools are the third route. We have not tested them.

Quality checklist

  • Output is 1080x1920, 9:16, H.264 video with AAC audio.
  • Captions and text sit inside the safe zone. The numbers we use are on Shorts, Reels and TikTok safe zone overlay.
  • The subject stays in frame in every shot, including after a cut.
  • The first second shows something that makes sense without sound, and captions carry the hook.
  • Loudness matches your other videos, and music does not cover speech.
  • The last frame is not cut off, and the loop point (if any) does not flash.
  • You watched it on a phone.

Known problems

Safe zones differ by platform and device. Google's published vertical ad zone on a 1080x1920 canvas leaves 288 px at the top, 672 px at the bottom, 48 px on the left and 192 px on the right, and Google notes that device screens vary and nothing is guaranteed. YouTube publishes no separate figure for organic Shorts. For Meta Reels, third-party guides report about 14 percent at the top, 35 percent at the bottom and 6 percent on each side, but we could not read Meta's own page. TikTok offers downloadable files and no fixed pixel zone.

Cropping a 16:9 video with burned-in captions cuts the captions off. Add captions after the reframe.

Face tracking can swing to the wrong person or jitter. Re-encoding the same file several times adds visible damage, so cut from the original, not from an earlier export.

Quick transitions and flashes can be hard on viewers who are sensitive to flashing. Keep flashes few and slow.

Tools and effects for this job

Shorts, Reels and TikTok safe zone overlay

Download a transparent safe zone overlay for YouTube Shorts, Instagram Reels and TikTok, with each zone's source and the date we checked it.

Kinetic typography slam

Heavy headlines that slam in on the beat, for hooks, ads and Shorts.

We turned a wide film clip into a vertical Short with free tools and checked the safe zone

We ran the Shorts and Reels job on three 60-second clips: tighten, find shots, crop to 9:16, caption in the safe zone, level sound. Timings and failures.

Pages on this site for vertical video: a safe-zone tool, a command builder and effects that suit short clips.

Reviewed picks from other creators

Reframing and short-video tools from other creators. Every card is a draft until the owner reviews it.

Source page screenshot, captured 2026-11-07

PySceneDetect

By Breakthrough on GitHub

Python tool that finds scene changes in video. The catch is that detection thresholds need tuning for each type of footage.

Use it for: Finds cut points in long footage before making shorts.

Licence, price and checks
  • Licence BSD-3-Clause
  • Reuse keep the licence notice
  • Library check works, Oct 10, 2026

Source reviewed on Oct 10, 2026. Not tested by us.

Source page screenshot, captured 2026-11-07

MediaPipe

By google on GitHub

Google's library for tracking faces and bodies. The catch is that it is a building block, not a ready editing tool, and its auto reframe example was not checked.

Use it for: Supplies face tracking for reframing wide video to vertical.

Licence, price and checks
  • Licence Apache-2.0
  • Reuse keep the licence notice
  • Library check works, Oct 10, 2026

Source reviewed on Oct 10, 2026. Not tested by us.

Source page screenshot, captured 2026-11-07

AI YouTube Shorts Generator

By Anil-matcha on GitHub

Open source generator that makes YouTube Shorts from longer videos. The catch is that the README pushes the author's paid API and clip quality depends on the language model.

Use it for: Makes shorts from a longer video automatically.

Licence, price and checks
  • Licence MIT
  • Reuse keep the licence notice
  • Library check likely works, Oct 10, 2026

Source reviewed on Oct 10, 2026. Not tested by us.

For teams that make this every week

If your team turns long footage into Shorts and Reels every week, tell us about the job, the stack and the step that slows you down.

If your team makes this kind of video every week, tell us the job, the tools you use, how many you make a week and where it gets stuck: contact us. We read every message; there is no form to fill in and no account.

Page checked on Nov 4, 2026.