Render host-side reference versions of custom effects against ordinary WAV
files before flashing a pedal. A .ZDL contains TI C674x machine code, so the
desktop cannot execute arbitrary release binaries directly. Each supported
effect has a small renderer that mirrors its pedal DSP formulas.
Install the Python dependencies:
python3 -m pip install numpy scipy soundfile
Show release-source effects and preview coverage:
python3 tools/audio_preview/preview.py list
Render TapeEcho4 using manifest defaults:
python3 tools/audio_preview/preview.py render tapeecho4 input.wav preview.wav
Override pedal-style parameter values:
python3 tools/audio_preview/preview.py render tapeecho4 input.wav preview.wav \
--set tempo=105 --set feed=82 --set flutter=48 --set wow=42 \
--set wear=66 --set drive=52 --set spring=0 --set mix=70
Render every effect that currently has a desktop adapter:
python3 tools/audio_preview/preview.py render-all input.wav preview-output/
The CLI deliberately lists effects without adapters as adapter needed.
Desktop renderers are listening tools, not substitutes for hardware tests:
they do not exercise the ZDL loader, runtime ABI, edit handlers, or pedal CPU
budget.
TapeEcho4’s desktop spring path is intentionally a listening target rather
than a mirror of the current ZDL fallback. It combines the first 744 ms of the
measured Galaxy spring capture with an eight-line FDN tail using 0.92
feedback, 0.55 damping, and a 320 ms equal-power handoff. The longer model
still needs a bounded pedal-safe implementation.
render_release_samples.py adds samples for the eleven releases not covered by
our original demo set. It preprocesses their current C sources, replaces the
firmware pointer prologue with host-owned memory, and runs the unchanged audio
body in eight-sample stereo blocks. Build the release pack first to generate its
parameter headers. Requires cc, NumPy and SoundFile; the shared-library output
currently targets macOS.
python3 tools/audio_preview/render_release_samples.py
python3 tools/audio_preview/check_samples.py
The script verifies Mix=0 preserves stereo input for all new samples with a Mix
control. Dustbox has Power instead. Sources are attenuated to 25%; rendered
outputs are only reduced if their peak exceeds 0.95. Settings, C source hashes
and gain are recorded in previews/audio/release_samples.json. The five-second
tails end with a 100 ms fade. These are host renders, not hardware recordings,
and do not test loader behavior, pedal CPU limits or converters.
The eleven earlier Python demos remain unchanged. Their original preset metadata
is unavailable. validation.json records objective file checks for both groups;
it does not claim a listening review or equality with the pedal.