Audio Loop Seam Finder
Set roughly where an audio segment should begin and end. The tool looks for nearby rising zero crossings, compares short waveform windows around each boundary pair, and weighs the amplitude and slope discontinuity at the loop junction. Listen to the proposed loop yourself and save the decoded segment as WAV. It does not understand musical structure or guarantee a perfectly silent seam.
Key features
- Decode local WAV, MP3, OGG, FLAC, M4A and other files that your browser supports into mono or stereo PCM.
- Search ±0–250 ms around each approximate boundary for rising zero-crossing candidates.
- Combine normalized window correlation with endpoint-amplitude and slope differences.
- Preview the proposed region on repeat and compare its endpoint jump with the original boundaries.
- Export a 16-bit PCM WAV loop and a JSON report with boundaries and scores without changing the source.
How to use
- Choose a 1–90 second audio file or load the generated WAV example.
- Enter approximate start and end times, search radius and waveform-window length.
- Find the seam and review the proposed boundaries, correlation and endpoint jump.
- Listen to the region on repeat, then save the WAV or JSON report.
Use cases
- Narrow down a natural repeat junction within a background-music cue.
- Compare candidate cuts near a hand-picked sound-effect region.
- Keep the candidate boundaries and search settings as JSON for another editing session.
Frequently asked questions
How does the seam search work?
It gathers up to 16 negative-to-positive crossings near each approximate boundary. For every valid pair it compares centered waveform windows, the last-to-first sample jump and slope difference. If there is no crossing, it falls back to low-amplitude points.
Does a large click-reduction percentage mean a perfect loop?
No. The percentage describes the absolute difference between just two boundary samples. Hearing also depends on slope, the surrounding milliseconds, rhythm and reverb. This is a candidate to audition, not a perceptual quality score.
Does it crossfade or beat-match automatically?
No. The WAV copies decoded frames from the recommended start through the frame before the recommended end as 16-bit PCM. It does not time-stretch, detect beats, crossfade or analyze musical sections.
Which file formats and limits apply?
Files may be at most 8 MiB and must be decodable by the current browser. Decoded audio must be 1–90 seconds, mono or stereo, 8–96 kHz and at most eight million frames. Browser codec support varies, and long compressed files can use memory while decoding.
Is my audio sent to a server?
No. The file is decoded and compared in this page's browser memory. JSON contains the file name, length, settings and boundary metrics but no audio samples. Only a WAV you explicitly save is downloaded locally.
Does the WAV keep the source audio quality?
The browser-decoded PCM segment is written as 16-bit integer WAV. A 24-bit, 32-bit float or lossy source may change precision or format, and samples outside ±1 are clipped. The rate and channel count follow the browser's decoded buffer.
Privacy
Selected audio is decoded and compared only in local browser memory, with no upload to a server API or remote analysis. The JSON report contains no audio samples. Playback and temporary download URLs are cleaned up when the input is cleared or the page closes.
Comments & questions