What we tested
We generated three mono 16-bit PCM WAV files at 22,050 Hz. Each file contained the same 440 Hz A4 tone, beginning at 0.250 seconds, lasting 1.800 seconds, and reaching the same 0.72 peak amplitude. Only the linear fade-in changed: 0.01, 0.30, or 0.90 seconds.
Every file used the same 0.02-second release and 0.25 seconds of silence after the tone. We ran all three through the current local converter at its default note-sensitivity and minimum-length settings, then inspected the exported MIDI files.
| Fade-in | First MIDI start | Events | MIDI end | File size |
|---|---|---|---|---|
| 0.01 s | 0.244 s | 1 × A4 | 2.068 s | 59 bytes |
| 0.30 s | 0.267 s | 1 × A4 | 2.068 s | 59 bytes |
| 0.90 s | 0.278 s | 2 × A4 | 2.068 s | 67 bytes |
What changed in the MIDI
The 0.30-second attack moved the detected start 0.023 seconds later than the 0.01-second attack. Extending the fade to 0.90 seconds moved the first detected start another 0.011 seconds, but it also split the sustained source tone into a 0.104-second A4 followed immediately by a 1.684-second A4.
All three MIDI files still ended at 2.068 seconds and used one musical track. The browser returned HTTP 200 for every run and reported no console errors. This test therefore shows a small timing shift plus an event-boundary change, not a lost pitch.
How to handle a soft attack
- Inspect the first note boundary. A softly emerging note may start a little later than the waveform's designed start.
- Merge adjacent duplicates. If one held pitch becomes two touching notes, join them in your DAW or piano roll.
- Keep the original audio nearby. Use the waveform or your ear when the precise entrance matters.
- Test the controls on your own source. Detection can change with pitch, tone color, noise, and minimum-note settings.
What this test does not prove
It tests one synthetic A4 at one peak level with three linear attacks. It does not cover bowed strings, breathy instruments, reverb, compression, mixed recordings, direct YouTube-link input, or other converters. It also does not establish a universal fade-in threshold.
For empty audio before a note, see the leading-silence timing test. For note tails, compare the trailing-silence test and echo test.