tannhatcms/jarvis/ams-audio-pro

Overview

Audio is half the experience and the first thing audiences notice when it's wrong — viewers forgive shaky video but not bad sound. This skill covers practical audio: recording clean source audio, editing dialogue, mixing basics (levels, EQ, compression), format choices, and loudness standards for different platforms.

When to use

• Recording voiceovers, podcasts, or interviews

• Cleaning up noisy or uneven audio

• Editing and mixing audio for video

• Choosing audio formats and export settings

• Meeting loudness standards for platforms or broadcast

Core concepts

• - - Source quality is everything. No plugin fixes a bad recording. Get the

mic close (6-12 inches for voice), record in a quiet treated space, monitor with headphones, and record 10 seconds of room tone for noise profiling.

• - - The signal chain. Record clean → edit (cut mistakes, tighten pauses) →

repair (noise reduction, de-ess, de-click) → mix (levels, EQ, compression) → master (loudness, limiting). In that order — mixing before editing wastes effort.

• - - Levels and headroom. Record peaks around -12 to -6 dBFS (never hitting

0). Mix dialogue to sit consistently; final masters peak at -1 to -2 dBTP (true peak) to survive lossy encoding.

• - - EQ: subtractive first. Cut problem frequencies (rumble below 80Hz on

voice, harshness 2-5kHz) before boosting anything. High-pass filter almost everything that isn't bass — muddiness lives in the low end.

• - - Compression gently. Voice: 2-4 dB of gain reduction evens out dynamics

transparently. Over-compression sounds lifeless and pumps. Learn to hear the threshold — if you can hear it working, it's too much.

• - - Loudness standards. Podcasts: -16 LUFS stereo (-19 mono). YouTube: -14

LUFS. Broadcast: -24 LUFS (US) / -23 (EU). Spotify: -14 LUFS. Master to the target — platforms turn down (not up) non-compliant audio.

Practical workflow

1. 1. 1. Plan the recording. Script or outline, quiet location (soft

furnishings kill reflections), mic choice and placement, test recording with playback check. Record room tone.

1. 2. 2. Record properly. 48kHz/24-bit for video work (44.1kHz/16-bit

minimum for audio-only), healthy levels with headroom, monitor on closed-back headphones, clap or slate for sync if multi-device.

1. 3. 3. Edit the performance. Remove mistakes and long pauses, tighten

breaths (don't remove all — breathing sounds human), smooth edits with tiny crossfades. Edit for pacing: podcast listeners forgive ums less than you think.

1. 4. 4. Repair. Noise reduction using the room-tone profile (gentle —

overuse creates underwater artifacts), de-ess harsh sibilance, de-click mouth noises, remove hums with notch filters.

1. 5. 5. Mix. Set dialogue level first (everything else supports it), EQ

subtractively, light compression, add music/SFX 15-20 dB under dialogue (ducking for voiceovers), check mono compatibility.

1. 6. 6. Master to spec. Loudness-normalize to the platform target,

true-peak limit at -1 dBTP, export in the required format (WAV master archived; AAC/MP3 for delivery).

1. 7. 7. QC on multiple systems. Headphones, laptop speakers, phone speaker,

car. If the mix only works on studio monitors, it doesn't work. Check the full listen for level jumps and artifacts.

Common pitfalls

• - - Recording too hot. Clipped audio is unfixable — distortion is forever.

Leave headroom; you can always turn up clean audio.

• - - Noisy environments. HVAC hum, room echo, street noise. Fix at the

source (turn things off, move rooms, get closer to the mic) — repair tools are damage control, not magic.

• - - Over-processing. Noise reduction artifacts, over-compressed

lifelessness, auto-tune sheen on podcasts. Every process should be inaudible in its operation.

• - - Music drowning dialogue. Background music mixed like a soundtrack

instead of a bed. Dialogue is king — everything else whispers.

• - - Ignoring loudness targets. Uploading a -24 LUFS podcast to a -16 LUFS

platform means listeners crank the volume, then get blasted by the next show. Normalize to spec.

• - - MP3 as a working format. Editing and re-exporting MP3s compounds

compression artifacts. Work in WAV/AIFF; export lossy only at the end.