Most of you don't listen to your own shorts
Audio is the half of short-form video everyone underinvests in. Voice cadence, the muted-feed reality, and the half-second nobody talks about.
I asked five creators this month to send me their last three shorts. Of fifteen videos, twelve had voice audio that was either compressed too hard, mixed too quiet against the music, or had pause patterns that sounded like nobody had read the script out loud before recording.
Audio is the underinvested half of short-form. Visuals get the attention because they're more legible — you can take screenshots of a frame, you can't screenshot a mix. So creators pour time into b-roll and overlays and color, and audio gets done in twelve minutes at the end.
I think this is wrong, and that it's costing most people 10-20% retention they could pick up basically for free.
The muted-feed problem
Roughly 60% of viewers in TikTok and Reels feeds have their volume off or very low when the video appears. That number gets quoted a lot and treated as a reason to add captions. Captions are good but they're not the actual lesson.
The actual lesson is about the other 40%. Those viewers decide whether to leave the volume up in the first ~0.5 seconds, based on what the audio sounds like. If your first half-second is quiet, mumbly, or has a music sting louder than the voice, those viewers will mute, and many will swipe.
The fix is dumb and most creators don't do it: the first half-second of audio should be the most confident, cleanest, most level-balanced half-second of the video. Not the loudest — the most certain. A clear voice saying a clear thing.
Pause patterns
I rerecorded one of my own videos last month with exactly the same script and a different pause pattern. The original had a pause after almost every clause. The recut had pauses only at idea boundaries — three pauses total in 28 seconds.
Same script. Same person. Same camera. Retention at second 15 went from 28% to 41%.
I don't fully understand why this works, but my guess is that breath-pauses feel like the speaker is processing, which signals "I'm not sure what I'm saying" to the viewer's pattern match for unscripted talking. Idea-pauses feel like the speaker is letting something land, which signals "this matters."
Read your script out loud before you record. If you find yourself pausing in places that aren't between ideas, fix the script — don't fix the recording.
Music
Most creators pick music that matches the mood of their video. That's wrong. Pick music that matches the energy of your audience's expectation for this niche on this platform.
A fitness video with melancholy lo-fi underneath gets muted because nothing else in the video matches it. The viewer's pattern match for "this is a fitness video" breaks. Confusion isn't curiosity in short-form. It's just confusion.
The boring answer: study what's working in your niche right now, pick from the same tonal pool, and don't try to be unique with music. Be unique with writing.
Native sound
Don't strip everything to silence under your VO. A faint room tone or low-mixed ambient track makes the voice sound present. Pure-silence VO sounds like a robocall.
You don't have to record the room sound on purpose. Just don't gate it to zero in post.
This is one of those areas where the thing I'm telling you to do is probably worth two of the things I told you not to. Audio is the dimension nobody competes on. You can win it with 30 extra minutes per video.