Expert Audio Quality Controller

Your agent for ensuring audio quality standards — covering quality assessment, artifact detection, noise analysis, and automated quality control workflows for audio production.

When to Use This Agent

Choose Expert Audio Quality Controller when:

• Assessing audio quality metrics (signal-to-noise ratio, dynamic range, frequency response)
• Detecting audio artifacts (clipping, distortion, noise, hum, dropouts)
• Building automated QC pipelines for audio production
• Setting quality standards for podcast, broadcast, or streaming audio
• Troubleshooting audio issues (echo, phase cancellation, codec artifacts)

Consider alternatives when:

• You need audio mixing and effects — use an Audio Mixer agent
• You need music composition — use a music generation agent
• You need video quality control — use a video processing agent

Quick Start

Copy
# .claude/agents/audio-qc.yml
name: Expert Audio Quality Controller
model: claude-sonnet
tools:
  - Read
  - Write
  - Edit
  - Bash
  - Glob
  - Grep
description: Audio quality control agent for quality assessment, artifact detection, and automated QC workflows

Example invocation:

claude "Create an automated audio QC script that checks for: clipping, silence gaps > 3 seconds, loudness deviation from -14 LUFS, and generates a quality report for each file in a directory"

Core Concepts

Audio Quality Metrics

Metric	Measurement	Target
Loudness (LUFS)	Integrated loudness per EBU R128	-14 to -16 LUFS
True Peak	Maximum sample value	< -1 dBTP
Loudness Range (LRA)	Dynamic range in LU	5-15 LU
Signal-to-Noise	Signal vs noise floor	> 50 dB
Sample Peak	Peak level in dBFS	< -0.3 dBFS

Quality Control Pipeline

Copy
# Automated QC Script Example
#!/bin/bash
for file in *.wav; do
  echo "=== QC Report: $file ==="

  # Measure loudness (EBU R128)
  ffmpeg -i "$file" -af loudnorm=print_format=json -f null - 2>&1 | \
    grep -A 12 "input_"

  # Detect clipping
  ffmpeg -i "$file" -af astats=metadata=1:reset=1 -f null - 2>&1 | \
    grep "Flat_factor\|Peak_level"

  # Check for silence gaps
  ffmpeg -i "$file" -af silencedetect=n=-40dB:d=3 -f null - 2>&1 | \
    grep "silence_"
done

Configuration

Parameter	Description	Default
loudness_standard	Target standard (ebu-r128, atsc-a85, custom)	ebu-r128
target_lufs	Target integrated loudness	-14
true_peak_limit	Maximum true peak level	-1 dBTP
silence_threshold	Silence detection level	-40 dB
clip_detection	Clipping detection sensitivity	strict

Best Practices

1. Measure loudness with EBU R128 for all broadcast content. Don't rely on peak levels for loudness assessment. Two files with the same peak level can sound drastically different in loudness. LUFS (Loudness Units Full Scale) measures perceived loudness, which is what listeners experience.

2. Check true peak, not just sample peak. Inter-sample peaks (between measured samples) can exceed 0 dBFS after D/A conversion. Use -af astats with true peak measurement or the ebur128 filter to detect these invisible peaks that cause distortion in playback.

3. Automate QC checks in your production pipeline. Don't rely on human ears to catch every issue. Build scripts that automatically check loudness, peak levels, silence gaps, and format compliance for every audio file before distribution.

4. Test audio on multiple playback devices. Audio that sounds fine on studio monitors may have issues on earbuds, car speakers, or phone speakers. Quality issues in the low and high frequency ranges are often masked by high-quality monitoring equipment.

5. Document your quality standards in a specification. Define acceptable ranges for every metric (loudness range, true peak, silence duration, sample rate, bit depth, codec) and share with all contributors. Subjective "sounds good" isn't a quality standard.

Common Issues

Loudness measurements vary between tools. Different tools may implement EBU R128 slightly differently, or measure at different stages (before/after codec). Standardize on one measurement tool (FFmpeg's loudnorm or ebur128 filter) and use it consistently across your pipeline.

Audio passes QC but sounds bad on streaming platforms. Streaming services (Spotify, Apple Music) apply their own normalization. Audio mastered at -8 LUFS gets turned down by 6 dB on Spotify (-14 LUFS target), potentially losing dynamic range. Master to the target platform's loudness standard.

Batch QC script takes too long on large file collections. FFmpeg's astats and loudnorm filters process the entire file to produce measurements. For initial screening, analyze a representative sample of each file (middle 60 seconds) before running full-file analysis on flagged items.