7 min read
How to make sound effects: sources, layers and the three part recipe
Almost every convincing effect is three sounds pretending to be one. Once you hear the structure you cannot unhear it.
A designed sound effect is rarely a single recording. It is a stack of layers, each doing one job, blended so the seams disappear. The structure is almost always the same.
The three part recipe
- Attack: the transient that sells the moment of contact. Short, bright, usually the loudest layer.
- Body: the mass and material. This is what tells you whether the thing is metal, wood, flesh or plastic.
- Tail: the decay and the space. Reverb, debris, resonance, the part that gives scale.
A door slam in a cathedral and a door slam in a car share an attack and differ almost entirely in tail. Build the three parts separately and you can retune the scale of any effect without re-recording it.
Where raw material comes from
Everything around you. Celery for bone breaks is a cliché because it works. A hand dryer pitched down two octaves becomes an engine. Ice in a glass becomes a spaceship door. The trick is to record objects for their texture, not for their identity.
- Record close, in a quiet room, with the mic no more than 30 cm away.
- Record more variations than you think you need, at several intensities.
- Leave two seconds of silence at head and tail so you can cut cleanly.
- Record at 96 kHz if you plan to pitch material down heavily.
Processing that does the heavy lifting
Pitch shifting changes perceived size more than anything else: down for mass, up for fragility. Time stretching turns an impact into a texture. Reversing a tail makes a rise. Distortion adds energy without volume. Use one obvious process per layer and keep the layers dry until the final blend.
Blending without mud
Each layer should own a frequency range. Filter the body out of the attack, roll the low end off everything except one layer, and check the result in mono. If it collapses in mono, two layers are fighting for the same band.
Prepare the files before you design
If your raw material came from a multitrack field recorder, split it into mono stems first and keep the take names. Working from named mono files instead of an eight channel poly file makes layering far faster, and the splitter here does it locally in your browser.