Captions, safe zones and the other boring details that decide performance
The unglamorous production details that make short-form video work: captions, safe zones, pacing and file formats - with the exact numbers we use.
A brilliant script can still underperform because the captions are hidden behind the like button. Here are the production details we check on every video, with the numbers we actually use.
Captions: word by word, big, high contrast
Many people watch short-form video with the sound off, so captions carry the message. What works best for us:
- A few words at a time. 4 words per caption on vertical video, up to 6 on horizontal. Long caption blocks get skimmed, not read.
- Highlight the spoken word. Colouring the current word (we use gold) keeps the eye moving with the voice.
- Timed to the audio, not estimated. We use word-level timestamps from the voice track, so every highlight lands on the right syllable.
- Thick outline or shadow. Captions must stay readable on any background.
- Break at punctuation. Never split a caption in the middle of a phrase if you can avoid it.
Safe zones on vertical video
TikTok, Reels and Shorts all draw their interface on top of your video: tabs at the top, like/comment/share buttons on the right, captions and the account name at the bottom. On a 1080×1920 frame, we keep anything important inside roughly:
| Edge | Keep clear |
|---|---|
| Top | ~160 px |
| Right | ~150 px |
| Bottom | ~400 px |
| Left | ~90 px |
That’s why our captions sit in the lower-middle of the frame, not at the very bottom.
YouTube banners and thumbnails have safe zones too
- Channel banner: 2560×1440, but only the centre 1546×423 is visible on every device. Put your logo and tagline there.
- Avatar: shown as a small circle - often under 100 pixels. Use a simple mark, not text.
- Thumbnail: 1280×720. The bottom-right corner is covered by the duration badge.
Pacing
- Cut on meaning. We change the shot at the start of each sentence, so the picture always matches what’s being said.
- Show things when they’re said. If the narrator says “roadmap”, the roadmap appears on that word - not two seconds before.
- Not too fast. Shots under about 1.5 seconds feel like flicker. We merge very short sentences into the next shot.
Audio
- Music under the voice. When the narrator speaks, the music drops well below the voice; on title cards it comes back up.
- Sound effects support motion. A soft pop when something appears, a whoosh on a transition. Subtle is the point.
- Speed without pitch. If narration feels slow, speeding it up slightly (with pitch preserved) works better than re-recording.
Delivery formats
| Use | Format |
|---|---|
| YouTube, landing page | 1920×1080, H.264, 30 fps |
| Shorts, Reels, TikTok | 1080×1920, H.264, 30 fps |
| LinkedIn, Facebook feed | 1080×1080 or 1080×1350 |
| Messaging apps | 720p copy, small enough to send |
Plus caption files (SRT) and a thumbnail for every video.
None of these details will make a video go viral on their own. But skip them, and even a great video will quietly underperform.