Bit Crusher (8-Bit Arcade) — User Guide

Digital degradation with true time-domain bit-depth quantization and sample-and-hold reduction, plus an alternate stepped spectral-shaping mode.

Author: Shai Cohen Affiliation: Department of Music, Bar-Ilan University, Israel Version: 0.6 (2026) License: MIT License Repo: Praat AudioTools

Contents:

What this does Quick start Processing modes Presets Parameters Output & visualization

What this does

Time Domain mode is the actual bit crusher. It reduces amplitude resolution with a selectable quantizer and can hold each sample value for several samples, creating the stair-step timing and aliasing associated with sample-rate reduction.

Stepped Spectral Shaping is a different effect. It applies a staircase of gain values across a selected frequency range. It does not quantize spectral magnitude, phase, or spectral word length.

Signal preservation: The script processes every channel independently and keeps the original channel count. Duration, stored sample rate, and the Sound object's original start time are preserved.

Quick start

In Praat, select exactly one Sound.
Run Bit_Crusher__8-Bit_Arcade_.praat.
Choose a preset, or leave Manual for direct control.
For classic bit crushing, use Time Domain mode and adjust Bit_depth and Sample_rate_reduction.
Choose the output-level behavior and click OK.

Starting point: Classic 8-bit uses 8-bit quantization with no sample hold. Heavy and Extreme progressively reduce both bit depth and temporal resolution. Telephone first band-limits the signal to 300–3400 Hz and then applies 8-bit / 4× hold processing.

Processing modes

Time Domain — bit crushing

The signal follows this path:

Optional band limit → amplitude quantization → sample-and-hold → output level stage

Bit_depth controls the amplitude grid. Lower values create larger amplitude steps and stronger quantization distortion.

Sample_rate_reduction is a whole-number hold factor. A value of 4 keeps one sample value for four original samples. The Sound object itself stays at its original sample rate; the displayed “effective SR” is the original rate divided by the hold factor.

Quantizer modes

Mode	Behavior
Symmetric, zero-preserving	Recommended. Symmetric positive/negative codes with an exact zero. Uses `2^bits − 1` levels, so silence remains exactly silent.
Signed PCM	Uses exactly `2^bits` signed codes including zero. The negative extreme reaches −1; the positive extreme is slightly below +1.
Legacy v0.3	Preserves the older response for compatibility. It uses `2 × 2^bits + 1` levels and also contains zero.

1-bit note: In the recommended symmetric mode, the minimum scale is clamped so 1-bit and 2-bit both use the same three levels: −1, 0, +1.

Stepped Spectral Shaping

The sound is transformed to a Spectrum. Inside Lower_frequency … Upper_frequency, each FFT bin is multiplied by a gain determined by its position in that range. The gain changes in discrete steps from 0 to 1.

frequency position → staircase gain with Quantization_steps + 1 levels
outside selected range → × Outside_range_multiplier

The same gain is applied to the real and imaginary components, so the spectral phase is retained. FFT zero-padding is trimmed after resynthesis so the output returns to the input duration.

Presets

Preset	Mode	Bit depth	Hold	Band limit
Manual	Uses the form values	Form value	Form value	Form value
Classic 8-bit	Time Domain	8	1×	Off
Subtle (12-bit)	Time Domain	12	1×	Off
Heavy (4-bit)	Time Domain	4	2×	Off
Extreme (2-bit)	Time Domain	2	4×	Off
Telephone	Time Domain	8	4×	300–3400 Hz
Crunch (6-bit)	Time Domain	6	3×	Off

What presets do not override: Quantizer_mode, the output-level mode, Ceiling_peak, visualization, and playback remain under user control. All built-in presets select Time Domain mode, so Quantization_steps has no audible role in those preset runs.

Parameters

Parameter	Default	Meaning
Preset	Manual	Selects Manual or one of six built-in Time Domain settings.
Mode	Time Domain	Actual bit crushing or stepped spectral gain shaping.
Bit_depth	8	Whole number from 1 to 24. Used only in Time Domain mode.
Sample_rate_reduction	1	Whole-number sample-and-hold factor. 1 = no hold reduction.
Quantizer_mode	Symmetric	Selects the amplitude-code mapping used by Time Domain mode.
Band_limit_before_crushing	Off	Applies a Hann band-pass before Time Domain quantization. Uses Lower/Upper frequency.
Lower_frequency	200 Hz	Lower edge for spectral shaping, or for the optional pre-crush band limit.
Upper_frequency	3000 Hz	Upper edge; values above Nyquist are clamped to Nyquist.
Quantization_steps	2	Whole number ≥1. In spectral mode this produces `steps + 1` staircase gain levels.
Outside_range_multiplier	0.5	Gain applied outside the selected range in spectral mode. Must be ≥0.
Output_level_mode	None	Leave level unchanged, attenuate only above a ceiling, or normalize every non-silent result to the ceiling.
Ceiling_peak	0.99	Target used by Safety ceiling or Peak normalize. Valid range: >0 to 1.
Draw_visualization	On	Draws waveform comparison, zoom, and summary.
Play_after_processing	On	Plays the final Sound after processing.

Output & visualization

Name: <source>_crushed_<preset>.
Duration: preserved.
Sample rate: preserved.
Channels: every source channel is processed independently and retained.
Time domain: the original Sound start time is restored after processing.

Output level

Mode	Behavior
None	No automatic level change.
Safety ceiling	Attenuates only when the current peak exceeds Ceiling_peak; quieter results are not boosted.
Peak normalize	Every non-silent result is scaled to Ceiling_peak, up or down.

Level warning: With no peak normalization, very low-bit processing can exceed ±1. The script reports this because saving such a result to integer PCM may clip.

Visualization

The Picture window shows:

the complete original and processed waveforms on a shared amplitude scale;
a side-by-side zoom of the first up-to-50 ms, also on the same shared scale;
a summary of the active mode, quantization or spectral settings, input/output peaks, output-level action, and channel count.

The visualization uses the final signal after the selected output-level stage.