Audio Transient Slicer
Turn a short recording of separate drum or impact sounds into a reviewable sample bank. The browser measures five-millisecond RMS blocks and proposes an onset when energy rises above your threshold and recent blocks. A minimum interval suppresses nearby repeated triggers. Set preroll and tail length, then inspect or adjust each start/end time before listening and saving individual 16-bit WAV files or one ZIP. Detection is an aid: quiet attacks, overlapping hits and background noise can require manual correction.
Key features
- Decode a local browser-supported audio file; the source stays in browser memory.
- Detect rising-energy attacks with adjustable dBFS threshold and minimum time between hits.
- Set preroll and maximum tail; proposed boundaries stop before the next detected onset.
- Review numeric start/end times, inspect a waveform with markers and listen to any proposed slice.
- Save individual 16-bit WAV samples, a CSV manifest or a ZIP containing all WAVs and the manifest.
How to use
- Choose a 1–90 second local audio file up to 8 MiB or load the generated three-hit WAV example.
- Set threshold, minimum interval, preroll and tail duration, then detect attacks.
- Inspect the waveform and edit each slice start/end time; play slices to check the cuts.
- Download individual WAV files, a CSV manifest or a ZIP sample bank.
Use cases
- Split separate snare, clap or percussion takes into reusable WAV one-shots.
- Find approximate impact positions in a short field recording and review them by ear.
- Keep source-relative onset and edited cut times alongside a sample bank.
Frequently asked questions
What counts as a transient?
The tool measures 5 ms RMS blocks averaged over all channels. A block must exceed the chosen dBFS threshold and be at least 1.8 times the strongest of the previous three blocks. Detected starts are then separated by the minimum interval. This simple energy-rise rule is not beat tracking or instrument classification.
Why is a hit missing or duplicated?
A soft attack may fall below the threshold; lowering it can also admit noise. A repeated impact within the minimum interval is suppressed. Overlapping impacts may not create a new energy rise. Adjust settings and review all cuts before export.
How are slice ends chosen?
Each slice starts slightly before its detected onset. Its end is the earlier of the next slice start, the chosen maximum tail length, or the source end. You can edit start/end times later. Overlapping edited slices are allowed and remain separate files.
Does export alter the sound?
The browser-decoded PCM frames in the reviewed range are written as 16-bit WAV at the decoded sample rate and channel count. There is no normalization, fade, time stretch or compression. Converting a 24-bit or lossy source to 16-bit WAV can change precision.
What are the limits?
The encoded file is at most 8 MiB. Decoded audio must be 1–90 seconds, mono/stereo, 8–96 kHz and at most eight million frames. At most 48 attacks are accepted; exceeding that shows an error instead of silently discarding slices. Browser codec support varies.
Is the recording sent to a server?
No. Audio decoding, onset detection, playback and ZIP creation take place in the browser. The manifest contains the source filename and timing, but no audio samples. The WAV files contain the reviewed slices.
Privacy
Selected audio is decoded, analyzed and packed only in browser memory. The page sends no file to a server API. Clearing the page releases the in-memory audio and playback; downloaded files remain on your device.
Comments & questions