-
Notifications
You must be signed in to change notification settings - Fork 4
Understanding the Output
When you run sounddiff file_a.wav file_b.wav, you get a report with five sections. Here's what each one tells you.
Basic facts about both files: duration, sample rate, channels, bit depth, and format. sounddiff flags mismatches here because comparing files with different sample rates or channel counts may reduce accuracy.
What to look for:
- Duration changes (one file is longer/shorter)
- Sample rate mismatches (48kHz vs 44.1kHz)
- Channel count differences (mono vs stereo)
Three measurements based on international broadcast standards:
| Metric | What it measures | Unit |
|---|---|---|
| LUFS | Overall perceived loudness | dB (LUFS) |
| True Peak | Highest signal peak including inter-sample peaks | dBTP |
| Loudness Range (LRA) | Dynamic range of the audio | LU |
The delta column shows how much file B differs from file A.
See What is Loudness? for a full explanation of these concepts.
Energy levels across frequency bands:
| Band | Frequency Range | What lives here |
|---|---|---|
| Sub | 20-60 Hz | Rumble, room noise |
| Bass | 60-250 Hz | Bass instruments, warmth |
| Low-mid | 250-500 Hz | Body of voices and instruments |
| Mid | 500-2000 Hz | Clarity, presence |
| Upper-mid | 2000-4000 Hz | Articulation, edge |
| Presence | 4000-6000 Hz | Definition, sibilance |
| Brilliance | 6000-20000 Hz | Air, sparkle, high harmonics |
The delta column shows energy changes per band. A positive delta means file B has more energy in that range. A negative delta means less.
See What is Spectral Content? for more.
How similar the two files are in structure and timing:
- Overall correlation: 1.0 means identical waveforms, 0.0 means completely unrelated
- Segments: Detected regions classified as similar, changed, added, or removed
See What is Temporal Structure? for details.
Potential problems found in either file:
- Clipping events: Samples hitting the maximum digital level, causing distortion
- Silence regions: Extended periods of silence that might indicate dropouts
See What are Clipping and Silence? for why these matter.
Throughout the report, deltas describe the difference between file A (reference) and file B (comparison):
- Positive delta: File B is higher/louder/more than file A
- Negative delta: File B is lower/quieter/less than file A
- Zero delta: No measurable difference
The reference file (file A) is always the first argument. Think of it as: "what changed from A to B?"