<!-- skill: Sound Design | version: 1.0.0 | description: Layering, EQ, reverb and loudness for rich soundscapes | category: Video & Design -->
## Instructions
- Layer with 3-layer rule: Body (low weight) + Character (mid texture) + Shine (high detail) for every major SFX.
- Record/procure at 96kHz/24-bit when pitching down, keep originals untouched in `/01-Field` folder.
- EQ carve: high-pass rumble under 30-80Hz, notch resonances, boost presence 2-5kHz only on hero layer.
- Use reverb sends (not inserts): Short room 0.6s + Long hall 2.5s buses, pre-delay 20-40ms for dialogue clarity.
- Pan deliberately: LCR for film hits, stereo width under 120% mono-compatible, check mono fold-down before print.
- Design dynamics: transient shaper attack +30% on hits, sidechain music -3dB under dialogue with 100ms release.
- Build whooshes: reversed cymbal + filtered noise sweep 200Hz-8kHz over hit duration, tail 0.5s past cut.
- Source ethically: Freesound (CC check), BBC Rewind/SFX archive, record foley for footsteps/cloth/props close-mic.
- Edit clean: 5ms fades on every region, de-click pitched assets, consolidate with handles 500ms for mix stage.
- Mix loudness: YouTube -14 LUFS / -1dBTP, Podcast -16 LUFS stereo, Film -24 LKFS, keep peaks under -1dBTP.
- Organize sessions: color Body blue/Character green/Shine yellow, name `EXP_impact_body_01`, buses `SFX/MUS/DX/VERB`.
- Reference on 3 systems: studio monitors + headphones + phone speaker, fix translation issues at -20dB quiet listen.
## Examples
- Cinematic hit: Sub sine 55Hz + metal clang pitched -7st + gravel crunch top, 0.8s tail hall.
- Forest bed: Wind loop + 3 bird spots panned L/C/R + leaf rustle gusts automated every 20s.
- UI click: 2ms tick 3kHz + soft wood knock, -14 LUFS, 150ms total with 10ms fade out.
## Workflows
### 1. Asset to layered cue
1. Spotting list with timecodes + emotional verbs per cue.
2. Gather 3 layers per cue, pitch/time-align to picture hit.
3. EQ carve + send reverbs, automate swells to cut.
4. Bounce stem (SFX only) + full mix, loudness-check.
### 2. Ambience build
1. Lay continuous bed loop with crossfades 2s.
2. Add sparse one-shots on separate tracks.
3. Automate filter/gain for movement, print 2min loopable.
## Templates
```text
Track layout: DX-01 | SFX-body | SFX-char | SFX-shine | AMB-bed | AMB-spots | MUS | VERB-room | VERB-hall | MASTER
Naming: PROJ_cue##_layer_version (e.g., TRAILER_hit_body_v02.wav)
Loudness targets: YT -14 LUFS/-1dBTP | Podcast -16 LUFS | Short -14 to -12 LUFS | TruePeak ceiling -1.0dBTP
Spotting row: 00:42 | Door slam + tension riser | Layers: wood + metal + air | Reverb: room 0.6s | Note: duck music
Sourcing log: filename | source (Freesound/BBC/field) | license | URL | credit line
```
## Tools
- DAWs: Reaper/Ableton/Logic, Valhalla Supermassive/Room, FabFilter Pro-Q3, iZotope RX.
- Libraries: BBC SFX Archive, Freesound.org, Sonniss GDC bundle, field recorder (H1n/H5).
- Meters: YouLean Loudness, Voxengo SPAN, mono/stereo correlation meter.
## Advanced
- Synthesize from scratch: FM metallic hits, granular stretched beds, convolution reverb from stairwell IR.
- Use mid/side EQ: mono bass under 120Hz, widen shine only above 6kHz to keep mix tight.
- Automate pitch + formant separately for creature vocals, layer human + animal + synth breath.
- Build Kontakt/DecentSampler instruments from 5 velocity layers for reusable signature sounds.
## Documentation
- Loudness specs: YouTube, Spotify, AES streaming tables; BBC SFX licensing terms.
- Theory: Haas panning, Fletcher-Munson curves, reverb pre-delay math.
- Internal: cue sheet, license tracker, stem archive with version notes.
### QA checklist
- [ ] 3 layers per hero cue (body/character/shine), fades 5ms, no clicks.
- [ ] EQ carved (HP 30-80Hz), mono bass under 120Hz, mono fold-down checked.
- [ ] Reverbs on sends (0.6s + 2.5s), pre-delay 20-40ms, dialogue unmasked.
- [ ] Loudness -14 LUFS YT / -16 podcast, true peak under -1dBTP, 3-system test.
- [ ] Licenses logged (Freesound/BBC/field), stems + full mix bounced + archived.
### Mix snippet
```text
Buses: DX -6dB / SFX -10dB / MUS -16dB ducked -3dB under DX / VERB sends -14dB.
Master: EQ > comp 1.5:1 > limiter -1dBTP > YouLean check > 48k/24b WAV + MP3.
Track colors: body blue / char green / shine yellow | Names PROJ_cue##_layer_v02.
```
- Pairs music-production for score ducking and audacity-cleanup for denoise.
- Complements podcast-audio loudness and voice-acting dialogue edit chains.
