What we tested
We generated three 22,050 Hz mono 16-bit PCM WAV files. Each contained the same isolated C4, E4, G4, and C5 sine tones, with each note lasting 0.75 seconds and a 0.35-second gap between notes. Only the opening silence changed: 0.25, 1.25, or 3.25 seconds.
Each file went through the current local converter at its default controls. Every browser run returned HTTP 200, completed without console errors, and exported a 105-byte MIDI file with one musical track and six note events.
| Opening silence | First source note | First MIDI note | Difference | Events |
|---|---|---|---|---|
| 0.25 s | 0.250 s | 0.255 s | +0.005 s | 6 |
| 1.25 s | 1.250 s | 1.266 s | +0.016 s | 6 |
| 3.25 s | 3.250 s | 3.252 s | +0.002 s | 6 |
What the result means
The exported MIDI did not move the first detected note to time zero. Adding one second of silence moved its start from 0.255 to 1.266 seconds; adding three seconds moved it to 3.252 seconds. In this test, the offset stayed within 0.016 seconds of the source note's designed start.
The pitch classes and total event count stayed the same in all three outputs. C4 and E4 were each split into a main event and a short edge fragment, while G4 and C5 remained single events. Small differences appeared in note boundaries and velocities, so the files were not byte-for-byte timing clones after subtracting the offset.
Should you trim silence before conversion?
- Keep it when the offset matters. If the audio should enter later on an existing timeline, the preserved opening can help retain that placement.
- Trim it for a time-zero draft. Remove unwanted dead air before conversion when you want the first detected note near the start of the MIDI clip.
- Check the imported clip position. A DAW can place the MIDI region on its own timeline, so distinguish note time inside the file from where the region is dropped.
- Review the first attack. Detection can begin a few milliseconds before or after the intended source boundary.
What this test does not prove
It compares three exact PCM files containing four synthetic notes. It does not set a universal timing tolerance, test trailing silence, measure a live YouTube-link workflow, or claim results for speech, complete mixes, tempo changes, compressed audio, or other converters.
If notes miss your DAW grid, see the tempo-alignment guide. For input encoding differences, compare the bit-depth test and sample-rate test.