Audio That Survives a Phone Speaker
Level the voice first, split the music so it can drop under speech, and check the mix on a phone speaker before cutting a frame of picture.
What it is for
An audio pass ordered for how people actually watch: voice levelled alone first, music brought in underneath it and cut into pieces so it can drop under speech, then two checks on a phone speaker. Do it before cutting picture. It prevents music that buries the narration and loudness jumps at every cut.
The skill
Most of your viewers are on a phone speaker in a room with other noise. Mix for that, not for your headphones. Do the audio pass before you cut picture. Fixing sound afterwards moves the timing you just spent an hour on. - Fix the recording, not the mix. Room echo does not come out. If the voice sounds like a bathroom, re-record closer to the microphone with something soft in the room; no tool removes reverb cleanly. - Level the voice first, alone, with music muted. Get one consistent voice level across the whole video before any other sound exists. - Bring music in far lower than feels right. Raise it until you notice it, then take it back down. Every consonant has to survive. - Split the music so it can duck. One long music clip holds one level, and a level that sits fine under a loud sentence buries a quiet one. Cut the music clip on the timeline at each point where speech starts and stops, then set each piece separately: down under the voice, back up in the gaps. - If you run a background-noise cleanup, listen to what it did. Cleaning too hard leaves a voice sounding like a radio underwater. Take less until it stops. - Match levels between clips recorded on different days before you worry about anything else. A loudness jump at a cut reads as a mistake even to people who cannot say why. Two checks before export: - Listen with the picture off. If you cannot follow the video by ear alone, neither can the viewer who looked away. - Listen on a phone at half volume with no headphones. If a word disappears, the mix is wrong; the listener is not. Choose instrumental music with no vocals and no drop - a lyric competes with your narration for the same attention and loses you both. And captions do not excuse bad audio: people leave a video that sounds bad before they read anything on it.
How to install it
1. Open the tool and start a new conversation, project or template. 2. Paste the text below wherever the tool keeps standing instructions — usually called system prompt, instructions, or persona. 3. Save it so it applies to every run rather than a single message.