Lesson 3 — Voice and Sound: The Half Everyone Skips
Mute any winning ad and watch it die. Audio is not garnish; it is half the video, and it is the half you can fix today.
- Audio is roughly half the video: voiceover, sound design and captions do most of the persuasive work.
- Raw, phone-quality voice often outperforms polished studio audio because it reads as platform-native rather than as an ad.
- Avatar and TTS pacing is driven by punctuation: short sentences and commas create natural rhythm, so lifeless delivery is a script problem.
- Layer sound as voice first, one or two diegetic effects, then music low and late.
Watch a winning ad with the sound off and it usually collapses. Voice, sound design and captions carry most of the meaning. Bad audio makes a video feel fake instantly, and it is where most AI ads quietly fall apart.
Counter-intuitive but reliable: a raw, phone-quality voice often outperforms a polished studio read on social. Studio gloss signals advertisement. Slight room noise and imperfect breath signal a real person talking to you.
Every TTS and avatar engine reads punctuation as timing. Short sentences and commas create rhythm. A long unbroken paragraph produces a flat monotone on any model. If your avatar sounds dead, fix the script, not the voice.
Clone one voice and reuse it across every variant so your brand sounds like one person. Direct it the way you would direct an actor, with pacing and emphasis baked into the writing. Keep settings fixed so variants stay consistent.
Three layers. Voice on top, always loudest. Then one or two real-world effects tied to what is on screen, a click, a pour, a zip. Then music, quiet and often starting late. Music that fights the voice loses you the viewer.
Action step. Take your script and re-punctuate it into short sentences with commas where you want a breath. Generate two reads, one clean and one rough. Add two diegetic effects, keep music quiet, and compare the two versions.
Loading the quiz…