It has to work silent and reward being unmuted

Why muted is the default, mechanically

Browsers and apps block audio that starts without a user action. This is not a stylistic preference of platform designers; it is a policy adopted after autoplaying sound became the most complained-about behaviour on the web, and it is now enforced at the level of the browser’s media engine. Audio requires a gesture: a tap, a click, an unmute.

The consequence is structural. Sound-on is an earned state, not a starting condition. Your piece begins silent, for everybody, every time, and the sound arrives only if something in the silent version persuaded someone to reach for it.

On top of that, a large share of viewing happens where sound is socially impossible: an office, a class, a shared room, a bus, next to a sleeping child.

The double build

The design rule follows directly and it is not “put captions on it”. It is:

1. The silent version must carry the whole meaning. Not a hint of it: the whole thing. If the piece is unintelligible without audio, it is unintelligible.
2. The sound version must add something, or being unmuted was a waste of the viewer’s decision.

Most work fails one or the other. A narration-driven film with subtitles bolted on fails the first: it is readable rather than watchable, because the meaning lives in words the viewer is now having to read while also looking at a picture. A silent-first piece built entirely from on-screen text fails the second: unmuting adds a music bed and nothing else, and the viewer has learned there is no reason to do it again.

The pieces that work put different information in each channel. The picture and on-screen text carry the argument. The sound carries texture, presence, rhythm, and the moments that only exist as sound: the machine starting, the room’s noise stopping, somebody laughing mid-sentence.

The phone speaker throws away your bass

Here is the specific technical fact that ruins more mixes than any other. A phone’s built-in speaker is a few millimetres across, and a driver that small cannot move enough air to reproduce low frequencies. In practice, output falls away sharply below roughly 500 Hz to 1 kHz, and there is effectively nothing under about 300 Hz.

So:

• A music bed whose weight is in the bass is, on a phone, a thin scratchy version of itself.
• A male voice recorded warm and close, with the low end that makes it sound rich on headphones, loses that body entirely and can end up sounding weaker than a thinner recording would have.
• A rumble, a heartbeat, a low drone: inaudible. The dramatic effect you mixed is not there.

The fix is not clever processing. It is checking on a phone speaker, held at arm’s length, in a room with other noise in it. Free tools do the rest: Audacity for a high-pass filter to clear out energy nobody will hear anyway, and any editor’s loudness meter to keep the level consistent.

Loudness, and why your piece is quieter than the next one

Platforms normalise loudness, measured in LUFS, so that one video is not dramatically louder than another. Targets vary and change, commonly landing somewhere around -14 LUFS for streaming. A piece mastered much louder gets turned down, and everything you crushed to make it loud stays crushed: you lose the dynamics and gain nothing.

A piece mastered much quieter gets turned up, which is usually fine, but a piece with a quiet voice and loud music gets normalised as a whole and the voice stays buried.

The practical rule: mix the voice to sit clearly above everything else, and leave the overall loudness alone. The platform will handle the rest, and the one thing it cannot fix is a balance problem inside your mix.

The first second, silent

The opening has to answer what is this and is it for me with the sound off. That means motion, a face, a legible few words, or a striking frame: something that works as a picture.

A common and avoidable failure is opening on a person starting to speak. Silent, that is a stranger moving their mouth: no information, and the viewer leaves before the sentence that would have hooked them finishes.

The limitation

Designing for silence pushes work towards on-screen text, and on-screen text competes with the picture for the same eyes. A viewer reading cannot simultaneously study the image, and a piece that solves silence by covering itself in text has traded one failure for another.

The resolution is a matter of sequencing rather than volume: put text where the picture is simple, and let the picture carry the moments that need looking at. That is an editing decision made at the structure stage, which is why sound-off design belongs at the planning table and not at the export dialogue.