How short a clip can still be detected?
Aureal repeats a frame about once per second, and detection needs enough of that frame to rebuild the codeword. On the named 12 s fixture, 1.8 s excerpts recovered at 49 of 93 start offsets and 1.0 s excerpts at about half; anything shorter than one frame (about 1 s at 44.1 kHz) never decoded. A short clip is not reliably traceable.
How Aureal handles it
More repetitions generally mean more integration and a steadier decision, which is why a full track is easier than a snippet. The circular folding path is what gives sub-2-second crops a chance: fragments on either side of a frame boundary are wrapped into their symbol slots instead of being discarded. That helps only when a complete 48-symbol codeword can be assembled. Below one frame there is not enough material, and around 1.8 s the result depends on where the cut falls, so an offset sweep is the honest test.
Common failure mode
“A 0.5 s snippet is enough to support a leak claim.”
Why it fails: Half a second is below one frame at 44.1 kHz, so there is no complete codeword to decode. Even at 1.8 s the local fixture recovered only about half the offsets, and a lossy encode on top can remove the marginal cases.
Verified example
Sweep the offset instead of testing one lucky clip:
ffmpeg -ss 3.0 -t 1.8 -i marked.wav -c copy clip-a.wav
ffmpeg -ss 7.4 -t 1.8 -i marked.wav -c copy clip-b.wav
node aureal-watermark.cjs detect clip-a.wav --id 883921
node aureal-watermark.cjs detect clip-b.wav --id 883921
Record each duration and offset. A single pass is not evidence that short clips are reliable.Example only. Validate against your own audio, codec and processing chain.
Scope and limits
- Anything shorter than one frame (~1 s at 44.1 kHz) is never recovered.
- At 1.8 s, about half the offsets recovered on the named fixture (49/93).
- The result depends on the start offset, source audio and codec.
- More frames help, but do not guarantee a specific clip.
Frequently asked questions
Can I detect a 5 s clip?
Usually there are several frames to integrate, so the odds are better than at 1.8 s. It is still a test on the actual file: codec, noise and the cut offset all matter.
Why did a 2 s clip fail when a 1.8 s clip passed?
Recovery depends on which fragments of the codeword the cut leaves and on the local signal, not on duration alone. The fixture's 49/93 result is a warning against treating any short duration as safe.
Related topics
Test the actual delivery chain. A fixture result is a starting point, not a guarantee for your material.
Open the Studio →