Slicer

Parameters

[Parameters not yet documented — likely declared in source outside register_animatable(). Add an override in _overrides/ if needed.]

Additional controls

Mode — How slice boundaries are detected:

• Equal — divide the input into N equal-length pieces (set by slice_count). Quick, predictable; good for chopping a 1-bar loop into 16ths.
• Duration — fixed-length slices (set by slice_duration_ms). Use when you want consistent slice length regardless of source duration.
• Transient — onset detection. Slices land at attacks (drum hits, plosive starts). Tunable via transient_threshold, transient_lookahead_ms, transient_window_ms. The MPC-style “auto-chop a drum loop” mode.
• Silence — slice at gaps below silence_threshold that last at least min_silence_ms. Useful for separating words in speech, or hits with audible decays between them.
• Manual — you place markers by hand. Each marker has a stable ID, so reordering preserves downstream Collage entry associations.

After running, you can drag any algorithm-placed marker to fine-tune — adjustments survive until you change parameters or hit “Reset to Algorithm”.

Slice Count — Number of equal-length slices to create. Only used in Equal mode. 2–64. For a typical 1-bar loop: 8 = eighth notes, 16 = sixteenths, 32 = thirty-second notes.

Slice Duration Ms — Fixed length per slice in milliseconds, 10–5000. Only used in Duration mode. Source is divided into chunks of this length; the last chunk takes whatever remains.

Transient Threshold — Detection sensitivity for onset detection, displayed inverted as “Sensitivity” (0–1) — higher value = more detected transients (more slices). Only used in Transient mode. Start at 0.3–0.7 and tune by ear; too high gets false positives on noise, too low misses softer hits.

Transient Lookahead Ms — How many ms before the detected transient peak the slice should start, 0–50. Only used in Transient mode. Capture pre-transient air / breath / pickup. Typical: 5–15 ms.

Transient Window Ms — Analysis window size for transient detection, 5–100 ms. Only used in Transient mode. Smaller = catches sharper transients (snare cracks); larger = smoother detection (only big peaks count). Typical: 15–30 ms.

Silence Threshold — Amplitude threshold below which audio counts as silence, 0.01–0.5 (linear, full-scale = 1.0). Only used in Silence mode. Lower = pickier (must be very quiet to count as a gap); higher = more lenient. Tune to your source’s noise floor.

Min Silence Ms — Minimum gap duration to trigger a slice, 10–1000 ms. Only used in Silence mode. Shorter = more slices on subtle pauses; longer = only big rests trigger. 50–200 ms is typical for speech.

Filter Enabled — Toggles the post-processing filter section. When on, slices that fail any of the filter_min_duration_ms / filter_max_duration_ms / filter_min_energy / filter_max_slices checks are removed from the output. Filtered slices show as struck-out on the waveform display.

Filter Min Duration Ms — Drop slices shorter than this, 0–1000 ms. Use to remove crumbs (very short fragments produced by aggressive transient detection on noisy material).

Filter Max Duration Ms — Drop slices longer than this, 100–10000 ms. Use to drop sustained / unwanted tail regions.

Filter Min Energy — Drop slices with RMS below this, 0–1 (normalized — 0.707 ≈ full-scale sine). Use to remove silent or near-silent slices that survived an over-sensitive Silence threshold.

Filter Max Slices — Hard cap on output count, 0 = unlimited. When set, only the first N (in time order) survive — the rest are filtered out. Useful when chaining into Slices Parser, which creates one output port per slice (capping prevents 50-port nightmares from over-detection).

About Slicer

Slicer is the entry point for K2K’s MPC-style sampling workflow: take a recording, chop it into discrete slices, then pipe to Collage (assemble into sequences) or Slices Parser (process each slice independently). Five detection modes cover the typical use cases — Equal/Duration for grid-based chopping, Transient for drum loops, Silence for speech / spaced hits, Manual for hand-placed markers. The waveform display is interactive: drag markers to tune, double-click to add (Manual mode), right-click to remove. Bypass passes the entire input as a single slice. Output 0 is the SliceContainer (the actual chopping); Output 1 is a passthrough of the input — handy for visualization and for chaining the un-sliced source into another branch. The Filter section runs after slicing and trims output by duration/energy/count constraints. Slices too short for FFT processing (< fft_size + hop_size samples) are auto-filtered regardless of the Filter toggle. Slice IDs are stable in Manual mode — Collage entries bound to slice ID 17 stay bound to it even if you reorder markers.

Generated 2026-05-05 from K2K_Dev@96730bdc by scripts/gen_lexique.py. Edit _intros/ or _overrides/, not this file.