Does the watermark survive a stereo-to-mono downmix?
Aureal spreads the watermark across the mid and side components. Because an in-phase stereo-to-mono downmix sums the channels, the mid layer gains roughly 3 dB of processing gain; the local phase-inversion fixture also decoded. Out-of-phase mixes, aggressive stereo widening and polarity tricks can still reduce or cancel a layer, so test the actual master rather than assuming mono is safe.
How Aureal handles it
For stereo input the embedder derives a mid watermark and a side watermark with separate key streams, then writes left and right as (mid + side)/√2 and (mid − side)/√2. Detection can score both the channel sum (mid) and the difference (side), which is why a simple polarity inversion does not automatically erase both layers. A standard L+R fold-down preserves the mid layer with about +3 dB; an L−R extraction preserves the side layer instead. The weak case is a mix where the two channels are substantially out of phase before the fold-down, because the sum can attenuate the very component that carries the mark.
Common failure mode
“Mono downmix is mathematically immune, so every mono file will decode.”
Why it fails: The +3 dB figure is processing gain for an in-phase sum, not immunity. If the channels are out of phase around the carrier, summing them attenuates the watermark, and heavy widening or all-pass processing can decorrelate the layers. The local phase-inversion fixture is evidence for that fixture, not a blanket guarantee.
Verified example
Test the exact fold-down, including a deliberately out-of-phase variant:
ffmpeg -i marked.wav -ac 1 mono.wav
node aureal-watermark.cjs detect mono.wav --id 883921
ffmpeg -i marked.wav -af "pan=mono|c0=0.5*c0-0.5*c1" side-only.wav
node aureal-watermark.cjs detect side-only.wav --id 883921
The first should favour the mid layer, the second the side layer. Keep the result with the master.Example only. Validate against your own audio, codec and processing chain.
Scope and limits
- In-phase downmix adds about +3 dB, not immunity.
- Out-of-phase content and stereo widening can attenuate a layer.
- The phase-inversion result is a local synthetic fixture.
- Detection scores the mono sum and the side difference, not every possible matrix.
Frequently asked questions
Does distributing in mono hurt recovery?
An in-phase L+R fold-down preserves the mid layer and adds about 3 dB of processing gain in the current detector. A mix with substantial out-of-phase energy is the case to test.
What about a vocal remover that subtracts the channels?
An L−R extraction removes the mid layer and leaves the side layer. The detector can score the difference, and the local phase-inversion fixture decoded, but treat that as a fixture result and test your processing chain.
Related topics
Test the actual delivery chain. A fixture result is a starting point, not a guarantee for your material.
Open the Studio →