AI Audio Detector with waveform slice analysis
Upload an audio file, mark one or more waveform slices, and review an explainable AI-audio likelihood report with per-slice evidence.
JolyMusic analyzes the selected audio locally in your browser. Load a file, use the waveform timeline to mark suspicious or representative passages, run the detector, then compare the global score with detailed per-slice findings.
Run a focused AI-audio inspection
- Upload an audio file and wait for the waveform timeline.
- Drag or resize colored slices over the passages to inspect.
- Run detection, compare every slice result, and export JSON when you need a record.
Detection report
From waveform slices to an explainable AI-audio report
Slice the evidence before judging it
Whole-file scores can hide local differences. Use the timeline, minimap, zoom controls, playback, and colored draggable slices to isolate intros, vocals, instrumental sections, edits, or suspicious moments before running the detector.
Read the detector as evidence, not a verdict machine
The report combines a likelihood score, verdict label, findings, and metrics such as dynamic range, spectral flatness, crest factor, RMS, peak level, clipping ratio, zero-crossing variance, and repetition. These signals help frame a listening decision; they are not forensic proof.
Compare multiple slices consistently
When several slices are present, each one receives its own detail card and score. This makes it easier to see whether a generated texture is isolated to one passage or repeated across the recording.
Build a reliable inspection habit
- Start with broad slices, then narrow the areas that look or sound unusual.
- Keep sensitivity stable when comparing slices from the same file.
- Use playback and zoom before interpreting the score.
- Export JSON for audits, teaching notes, or repeated comparisons.