How the audio
holds up under testing.

Perceptual quality

Watermark evaluation considers both listening quality and recovery. The following figures are the internally reported results from Anubis Protect's existing evaluation.

PEAQ ODG

+0.04

ViSQOL MOS-LQO

4.73 / 5

Signal-to-noise ratio

~38 dB

Noise-to-mask ratio

-27 dB

Reported results describe the evaluated material. They are not an independent certification or a guarantee for every recording.

The evaluation also includes controlled A/B/X listening comparisons. The listening samples let you compare original and watermarked recordings directly.

Attack and adversarial testing

Recovery tests apply transformations that can occur during distribution, or be used in attempts to disrupt a watermark.

Editing

Trimming, from milliseconds to multiple seconds; gain changes and clipping.

Compression

MP3, AAC and Opus, including repeated codec conversions.

Resampling & filtering

44.1 kHz and 32 kHz round trips; low-pass and high-pass filtering.

Pitch & timing

Current published ranges: up to ±5 semitones with speed preserved, and ±12% time stretching with pitch preserved.

Combined processing

Stacked transformations, noise injection and dynamic processing.

Distribution workflows

DAW exports, video uploads and audio recovered from streaming services.

Codec robustness

Lossy compression changes the signal carrying a watermark. The existing evaluation reports the following behaviour.

These are robustness test conditions, not a list of accepted upload formats. Recovery can vary with the codec settings, source material and processing chain.

Validation on your material

A useful evaluation includes the recordings and transformations in your own distribution workflow.

1. Include sparse recordings, dense masters and different production styles.
2. Compare the protected audio against the original under your normal listening conditions.
3. Apply the compression, edits and resampling used in your distribution chain.
4. Decode the resulting files and record recovery failures as well as successes.

For a technical assessment, contact us about the test conditions and evidence you need to review.

Validation of AI detection

Verify exposes spectral, temporal and phase evidence, with a segment timeline to show how it changes within a recording.

Assessment needs to account for false positives, false negatives, mixed human and synthetic material, and processing after generation. A result should be interpreted alongside production history and source material.