Scans audio for common technical and tonal problems. Returns a severity-coded list of detected issues with frequency locations and suggested fixes.
Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
| file_path | string | required | Path to audio file |
Example Output
Illustrative summary; MCP responses use structured JSON.
$ detect_problems vocals.wav
What the Numbers Mean
Severity levels (most to least serious):
- dealbreaker — Highest-priority detector flag. Inspect the source and delivery needs first; a severity label is not a listening test.
- significant — Likely audible and worth addressing.
- moderate — Noticeable on careful listening. Worth considering.
- minor — Technical issue but may not be audible. Optional fix.
Problem types:
- clipping — Samples at the configured amplitude threshold, checked on each channel. At the default threshold, integer PCM’s representable full-scale rails are also detected. A flagged rail is a reason to inspect the source; reducing gain cannot restore already clipped detail.
- dc_offset — Non-zero baseline. Reduces headroom and can cause clicks at edit points.
- inter_sample_peak — Reconstructed peaks exceed sample peaks by the configured overshoot threshold. Severity also considers true-peak level; a flag does not necessarily mean the peak exceeds 0 dBTP.
- noise_floor — Broadband noise floor above acceptable levels.
- snr — Signal-to-noise ratio below professional levels.
- hum — Power line interference at 50/60 Hz and harmonics.
- sibilance — Harsh “s” and “t” sounds, 5-10 kHz.
- mud — Excessive low-mid energy, 200-500 Hz, that obscures clarity.
- harshness — Aggressive upper-mid energy, 2-4 kHz, that causes listener fatigue.
- resonant_peak — Narrow spectral peaks that may warrant listening; the measurement does not establish their cause.
- lossy_codec — A sharp high-frequency shelf that may indicate prior lossy encoding. It is a heuristic, not proof of a file’s history.
There is no sensitivity parameter on this MCP tool. To change a detector threshold, use the documented analysis settings and report which defaults you changed.
Example Prompts
Quick check
Are there any problems with my vocal recording at vocals-take3.wav?
Mix problems
Check my mix for any technical issues before I send it to mastering
Sensitive scan
Run a sensitive problem detection on master.wav — I want to catch everything
Related Tools
- full_diagnostic — Includes problem detection plus all other analyses
- analyze_spectrum — Detailed frequency view to investigate detected mud/harshness
- analyze_loudness — Check if clipping is related to overlevel
- fix_audio — Automatically apply corrective processing for detected problems
Pro tip
Run problem detection early — before you start mixing. Fixing DC offset, hum, or sibilance at the source is always better than processing a mix that already has these issues baked in.