Sound design

Sound Design

Audiences forgive bad picture far more than bad sound. Muddy, inconsistent, or fatiguing audio reads as "amateur" instantly. The goal is a mix where every element is intelligible, the levels never surprise, and the sound serves the story.

Build in layers

Most video audio is four stems. Mix them as separate layers, then balance:

1. Dialogue / voiceover — almost always the priority. The viewer must understand every word.
2. Music — sets emotion and pace; ducks under dialogue.
3. Ambience / room tone — the bed of background sound that makes a space feel real and hides edits.
4. Sound effects (SFX / foley) — specific sounds tied to on-screen actions, plus designed effects (whooshes, impacts, risers).

Dialogue comes first

1. Clean it — denoise (gently — over-denoising creates underwater artifacts), remove hum (50/60Hz), de-ess harsh sibilance, and use a high-pass filter ~80–100Hz to cut rumble.
2. EQ for clarity — cut mud around 200–400Hz, add presence around 3–5kHz for intelligibility. Subtractive before additive.
3. Compress — even out the dynamics so quiet words and loud words sit close. Gentle ratio (2:1–4:1), catch the peaks.
4. Level consistently — every line/clip at a consistent perceived loudness before you add anything else.

If dialogue isn't clean and even, nothing layered on top will save it.

Room tone hides your edits

Every time you cut dialogue, you cut the background too — and silence between clips is jarring. Lay a continuous bed of room tone / ambience under the whole track. It glues edits together and makes cuts inaudible. Record 30 seconds of "silence" on location for exactly this; if you don't have it, match with a library ambience.

Music: serve, don't compete

• Duck the music under dialogue — automate it down 6–12dB whenever someone speaks (sidechain or manual). The viewer should never strain to hear words over a score.
• Cut music to picture — let musical hits land on visual beats, and edit the track so it resolves rather than fading mid-phrase.
• Match energy to the scene. Don't let a driving track flatten a quiet moment, or vice versa.
• Mind the licensing — note when a track needs clearance or is royalty-free/library.

Sound effects and atmosphere

• Sync foley to action — footsteps, cloth, clicks, impacts. Even subtle foley massively increases realism.
• Design transitions — whooshes and risers smooth cuts and build anticipation; impacts punctuate.
• Layer for depth — a "big" sound is usually several sounds stacked (e.g. an impact = low boom + mid thud + high crack).
• Use silence deliberately — dropping sound out before a hit makes the hit land harder.

Levels and loudness — hit the target

Mix to the platform's loudness standard so your video isn't quieter or louder than everything around it:

• YouTube / streaming — around −14 LUFS integrated.
• Podcast / spoken — −16 LUFS (mono often −19).
• Broadcast (TV) — −23 LUFS (EBU R128) / −24 LKFS (ATSC US).
• True peak — keep below −1 dBTP to avoid clipping on lossy playback.

Use a loudness meter, not just peak meters — perceived loudness ≠ peak level. Set dialogue as your anchor (e.g. dialogue averaging around −12 to −10 dBFS short-term), then build everything else relative to it.

Mix order

1. Edit and clean dialogue → 2. level dialogue → 3. lay ambience/room tone → 4. add and duck music → 5. add SFX/foley → 6. balance the full mix → 7. master to the loudness target → 8. check on small speakers (phone, laptop) where most people will actually hear it.

Output format

## Audio plan: <project> → <deliverable> (<LUFS target>)

### Dialogue chain
- Cleanup: <denoise / hum / HPF>
- EQ: <cuts/boosts>
- Compression: <ratio / target level>

### Layers
- Music: <track> — duck −<n>dB under VO, cut to <beats>
- Ambience: <bed> — continuous under whole track
- SFX/Foley: <key moments>

### Mastering
- Integrated: <LUFS>   True peak: <−1 dBTP>

### Cautions
- <noise floor, licensing, mono-compatibility, small-speaker check>

State the loudness target up front (it's deliverable-specific) and always recommend a final check on phone/laptop speakers — the mix that sounds great on monitors often loses the dialogue on a phone.