Show HN: I reverse engineered serum 2 and build an open source alternative

Tessera is an open-source reimplementation of Serum 2: a wavetable synth written from scratch in C++ on JUCE that loads and saves .SerumPreset files and uses Serum 2 content (wavetables, samples, presets) you already have installed. Its engine was rebuilt by measurement and binary analysis module-by-module until its audio output nulls against Serum 2: oscillators (wavetable, sample, multisample, granular, spectral, sub, noise), all 108 filter types, warps, 13 FX units, modulation matrix, LFOs, envelopes, arpeggiator and RNG. Quantitative validation is central: a null-test suite (~3,050 cases) reports 0 failures at 48 kHz (also at other rates), all 850 factory presets compare within 1 dB, median 1/3-octave band distance 0.00 dB, and a feature matrix lists 248 implemented features, 15 not applicable and none missing.

The project includes a Python test harness (null tests, preset sweep, blind ABX) that requires Serum 2 as the reference, and tools to probe which module causes differences. It builds on macOS (VST3, AU, standalone) using CMake, Ninja and Xcode tools; Windows and Linux ports are not yet provided. Tessera ships no proprietary content, is MIT-licensed (JUCE licensing constraints apply to binaries), and asks contributors to include null-test cases for DSP changes. Performance targets aim to be at most ~1.3× Serum's render time on factory presets.