ACULTUR8 · LOCAL SAMPLER PROTOTYPE

Find the hits. Keep the groove.

Load a WAV, preview slice detection, then choose which hits to use. Audio stays in this browser.

Browser review · no FM1 connection or upload

1 · Source audio

Mono or stereo PCM WAV, up to 30 seconds at 8–48 kHz. The original audio stays unchanged.

Download the demo WAV to pair with exported markers

2 · Slice profile

Hold, Decay and Level above are starting defaults for newly accepted slices. Existing slices keep their own settings. Envelope off preserves the full slice length; Level still applies. Full-band uses the whole signal. Other profiles filter detection only. Less-negative attack sensitivity values find fewer hits.

Maximum slice length limits how much audio is kept after each cut, including pre-roll. Set 0 to keep audio up to the next detected slice. Preview before applying; accepted edits stay unchanged until you use the proposed markers.

Detection view · find a useful frequency band

Inspect a short hit: click its position in the original waveform below, or set a region here. The spectrum shows the original audio in that region; the shaded area shows the detector band. A spectrum peak is not an instrument label.

Load audio to inspect its spectrum.

Band energy is measured as stereo-averaged RMS in 2 ms windows. Minimum level is absolute dBFS, separate from relative attack sensitivity.

▲ Magenta: detected attacks · Orange: proposed slice starts · Green: accepted active slices · Grey: inactive slices / peaks below minimum level. Dashed yellow line: minimum band level.

3 · Review slices

Load the demo or a WAV to begin.

Scroll horizontally through the slice buttons. Inactive slices remain editable and auditionable; group playback skips them and preserves the original gaps.

No slice selected

Copy/paste includes Active, Length, Hold, Decay and Level, keeping the destination start and label. Reset makes this slice Active and restores the current profile's length and sound defaults.

Length is the maximum source boundary. With Hold + Decay enabled, sound fades linearly to silence after Hold, stopping at the source boundary if reached first. Edge fades reduce clicks; no sustained sound is generated. Group trigger positions stay unchanged. Click Apply slice sound to commit these sound controls.

Save a review project to preserve every slice's sound settings and Active flag; reopen it with the same WAV. Marker export pairs with the original WAV in the existing workbench but contains cuts and timing only: Hold, Decay and Level are not applied there. No audio is embedded in JSON.

Play these sounds with the emulator: export a review project, then choose that JSON and the same original WAV in the browser emulator. It applies Active, Length, Hold, Decay and Level. Save/reopen in the emulator also keeps the original audio. This is browser playback; physical FM1 upload of these settings is not implemented.

How detection and editing work

Detection compares short-window energy against a slower envelope, then finds spaced onset peaks. A candidate must pass both relative attack sensitivity and minimum band RMS dBFS, then minimum spacing. Target frequency and bandwidth focus the analysis; stereo channel energies are combined without phase cancellation. Slice starts look backwards by the pre-roll and zero-crossing search range in the original audio. Stereo uses channel 1 crossings and favours low combined energy.

Magenta triangles show actual onset matches; orange cut lines can precede those attacks because of pre-roll. Green marks accepted active cuts and grey marks inactive cuts. Matches only reflect the most recent Preview detection. Changing detection settings clears that proposal, preserving accepted edits.

The spectrum is a 4096-point Hann-windowed average of original stereo channel powers over the inspected region. Levels use peak-sine dBFS; energy uses RMS dBFS, so a full-scale sine reads approximately 0 dBFS on the spectrum and −3 dBFS RMS. Log frequency axis, 20 Hz to Nyquist; resolution depends on source sample rate. Short hits and spectral leakage can spread peaks. Shading indicates the approximate half-power band of a second-order band-pass, not a brick-wall cutoff.

Hear detection band uses the same filter as analysis, without normalization, through the Preview level and short fades. Stop audio stops all audition. Full-band mode auditions the original. The original PCM is unchanged; filtering, minimum level and coloured matches do not isolate or remove other instruments from exported slices.

Boundary fields edit original source frames. Real recordings can need manual correction; this prototype does not separate instruments, decode REX, time-stretch, record USB audio, or change firmware. Custom profiles are stored for this browser origin; export profiles and review projects for portable copies.

Icons: Font Awesome Free 5.15.4 · CC BY 4.0 / license. Bundled locally for offline use.