## Instructions - Always work non-destructively: duplicate the raw track (Ctrl+D), mute the backup, edit the copy, export WAV master before any MP3. Set project rate 44100 or 48000 Hz before recording, snap-to off for breath edits. - Capture a real noise profile: select 1-2s of pure room tone (no breath/speech), Effect > Noise Removal > Get Profile, then apply Reduction 8-12 dB / Sensitivity 6.00 / Frequency smoothing 3-6. Never over-reduce past metallic artifacts. - Apply noise reduction in two light passes (6 dB + 6 dB) instead of one 18 dB smash for voice that stays natural, previewing sibilants each time. - Truncate silence surgically: Effect > Truncate Silence at -35 dB threshold, 0.3s min / 0.5s max silence, compress to 0.4s. Manually keep dramatic pauses; auto-truncate only breaths and dead air. - De-click mouth noises: zoom to click, Spectral Delete or 5-10ms silence patch, declicker threshold light. De-esser via EQ dip at 5-8 kHz if harsh, never full-band NR for mouth clicks. - Run the voice chain in order: Noise Reduction > High-pass 80 Hz (6 dB/oct) > Gentle EQ (cut 200-300 Hz mud, +2 dB presence 3-5 kHz) > Compressor 2.5:1, -18 dB threshold, 10ms attack, 100ms release > Limiter -3.5 dB ceiling. - Hit ACX audiobook spec exactly: RMS -23 to -18 dB, peaks under -3 dB, noise floor below -60 dB RMS. For podcasts target -16 LUFS stereo loudness normalize + -1 dBTP true peak instead. - Use Loudness Normalize last: Effect > Loudness Normalization > Perceived loudness -19 LUFS mono audiobook / -16 LUFS stereo podcast, then Limiter to catch peaks. Measure with Analyze > Measure RMS + ACX Check plugin. - Remove room echo cheaply: close-mic 4-6 inches, Noise Gate at -40 dB (attack 10ms, hold 200ms) only after NR, never gate before profiling noise. Re-record hollow rooms rather than over-processing. - Edit breaths like a human: reduce loud breaths -6 to -10 dB (Envelope tool), don't delete all breaths — zero-breath audio sounds robotic. Keep 0.2s room tone under every cut to avoid digital black gaps. - Export correctly: WAV 16/24-bit master archive, MP3 192kbps+ CBR for distribution, mono for audiobooks (smaller + ACX-friendly), stereo only if music/SFX panned. Embed ID3 (title/author/chapter). - QC on three systems: studio headphones, phone speaker, car. Listen at 1x for artifacts, check head/tail 1s room-tone padding, verify no clipping (View > Show Clipping red lines). ## Examples - Good: 10 dB two-pass NR from real room tone, truncate to 0.4s, -19 LUFS mono, -62 dB floor, passes ACX. Bad: 24 dB single-pass NR, underwater voice, -55 dB floor, rejected. - Good breath edit: loud inhale dropped -8 dB with envelope. Bad: all breaths deleted, choppy robotic cadence. - Good export: `ch03_master.wav` + `ch03_192.mp3` mono ID3-tagged. Bad: single 64kbps joint-stereo MP3 as only master. ## Workflows 1. Prep: duplicate track, profile 2s room tone, backup raw WAV. 2. Clean: two-pass NR > truncate silence > de-click > high-pass. 3. Polish: EQ > compressor > loudness normalize > limiter > ACX/RMS measure. 4. Ship: WAV master, MP3 distro, three-system QC, log settings per chapter. ## Templates - Audacity macro chain (Tools > Macros): - `NoiseReduction: reduction=10 sensitivity=6.0 smoothing=4` - `TruncateSilence: threshold=-35 min=0.3 max=0.5 compress=0.4` - `HighPass: freq=80 rolloff=6` - `EQ: cut -3dB@250Hz width=1oct, +2dB@4000Hz width=0.7oct` - `Compressor: threshold=-18 ratio=2.5 attack=0.01 release=0.1` - `LoudnessNorm: LUFS=-16 stereo / -19 mono` - `Limiter: ceiling=-3.5dB (ACX) / -1.0dBTP (podcast)` - Room-tone patch: `Generate 0.4s silence > paste ambient noise sample at -50dB under cuts` - ACX check targets: `RMS -23..-18dB | Peak < -3dB | Floor < -60dB` ## Tools - Audacity Noise Reduction, Truncate Silence, EQ, Compressor, Loudness Normalization + ACX Check plugin. - ffmpeg for loudness verify (`loudnorm` filter), SARAH file tools for chapter logs and export trackers. ## Advanced - Chain Macros batch: apply the full voice chain to a folder of chapters overnight via Tools > Apply Macro to Files. - Build per-mic profiles: separate NR profile + EQ per guest/mic, save as Macro presets, never reuse host profile on guest track. - Rescue clipped peaks: Effect > Clip Fix on isolated spikes only, then re-limit; re-record if more than 5 clips per minute. - Two-pass NR math: 8 dB pass one + 6 dB pass two with fresh preview each time; stop the moment sibilants turn metallic. - Truncate Silence tuning: audiobooks -40 dB / 0.5s max (preserve narration pacing), podcasts -35 dB / 0.4s (tighter energy). - EQ cheat sheet vocals: high-pass 80 Hz, cut 250 Hz -3 dB mud, +2 dB 4 kHz presence, low-pass 15 kHz hiss trim on cheap mics. - Compressor starting point: threshold -18 dB, ratio 2.5:1, attack 10ms, release 120ms, makeup to -19 LUFS before limiter. - ACX fail triage: floor fail = re-profile longer room tone + lighter NR; RMS fail = re-run Loudness Norm before limiter; peak fail = lower limiter ceiling. - Room-tone bed: paste 10s clean tone under entire chapter at -50 dB to mask edit-point silence gaps. - Punch-and-roll habit: record fixes in place (Shift+D) instead of patching separate takes — matches tone and saves edit time. - Backup chain: raw WAV + edited project + WAV master + MP3 distro, chapter-named, mirrored to second drive before delete. ## Quick Reference - Safety: duplicate track, 44.1/48kHz project rate, WAV master before any MP3. - NR: 1-2s room-tone profile, two light passes (8+6 dB), sensitivity 6.0, smoothing 3-6. - Truncate: -35 dB, 0.3s min / 0.5s max, compress to 0.4s; keep dramatic pauses manual. - Chain: NR > high-pass 80Hz > EQ (cut 250, +4k) > 2.5:1 comp > loudness > limiter. - ACX: RMS -23..-18 dB, peaks under -3 dB, floor below -60 dB; podcast -16 LUFS stereo. - Breaths: -6 to -10 dB envelope, never delete all; 0.2s room tone under every cut. - QC headphones + phone + car; 1s room-tone head/tail; mono MP3 192k for audiobooks. ## Documentation - Audacity Manual: Noise Reduction, Truncate Silence, Macros, Loudness Normalization (manual.audacityteam.org). - ACX audio submission requirements (audible.com/acx) — RMS/peak/noise-floor spec. - Podcast loudness: AES/EBU -16 LUFS stereo guidance. - ffmpeg loudnorm two-pass documentation for verification passes.