How my shorts always land at 15-20 seconds without me editing anything
Ask ten shorts creators how long a short should be and you’ll get ten guesses. I stopped guessing: my pipeline enforces a 15-20 second window automatically, on every video, without touching a timeline.
Why 15-20 seconds
The numbers that led me there, from my own animal shorts channel:
| Metric | Number | Source |
|---|---|---|
| Best video retention | 74.9% | YouTube Studio |
| Second best | 71.2% | YouTube Studio |
| First-second swipe-away | 59.9% | YouTube Studio |
| Videos shipped in this format | 60+ | channel dashboard |
Both top performers sit inside that window. This isn’t a universal law, and I won’t pretend it is: it’s the window that works for a fact-per-scene animal format. A storytelling channel may need triple that. The point is to pick a window from your own retention data, then enforce it.
The trick: adjust the voice, not the script
Scripts never come out the right length. The usual fix is rewriting or trimming clips. My pipeline does neither. It generates the narration first, measures the real audio duration, and if the total falls outside 15-20 seconds, it regenerates the voice at a different speaking rate until it fits. A slightly long script gets read a touch faster. A short one, a touch slower.
Because subtitles are positioned from the measured audio, not from estimates, everything stays in sync no matter which rate it picked.
Small print, honestly stated: there is a limit to how far you can push rate before narration sounds rushed. My pipeline stays inside a narrow band; if a script is so long it would need aggressive speed-up, that’s a script problem, and I cut text instead.
What you can use today
Even without a pipeline, you can steal the principle:
- Pull retention graphs for your last 10 videos and note where the drop-off cliff sits. That’s your window telling you where it is.
- Read your script aloud with a stopwatch before you edit. If it’s over your window, cut words now, not clips later.
- If you use any TTS tool, generate narration before touching the editor, and cut the video to the audio. Audio-first is the whole game.
If you want the pipeline version of this, the earlier posts in this series cover the free TTS setup and the CapCut draft trick it rides on.
Here are the receipts.