Hi Product Hunt I built SonicForge because every online tone generator is a slider and a play button, and I kept needing more than that. This came out of a combustion-acoustics experiment where the excitation had to be controlled: known, repeatable, comparable between runs. No tool I tried offered phase at all, and each one named its levels differently. The part I would point at first: two channels 180 degrees apart cancel to exact silence. That sounds obvious, but OscillatorNode in the Web Audio API has no phase parameter - you genuinely cannot start two oscillators at a chosen offset. SonicForge builds each waveform from its Fourier coefficients and rotates every harmonic, so the phase is baked into the wave table, and every channel is locked to the audio clock. The test suite asserts the residual is 0.00e+0. What is inside: 16 tone channels with level in dBFS and phase in degrees, seven noise colours, microphone room calibration that measures round-trip latency first, a scripting terminal that runs on the audio clock, a WebGL2 3D spectrogram, and hardware presets like speaker water eject and headphone burn-in. Things I decided not to do: - No Three.js. The 3D spectrogram is raw WebGL2, because a CDN dependency would break the offline guarantee. - No QR library. The encoder is written from ISO/IEC 18004. - No backend, no analytics, no account. The microphone opens only during a calibration measurement and is released in a finally block. There are 120 assertions plus a 70-check application self-test, and nothing is mocked: the noise tests measure the real spectrum of a generated buffer, and the integration tests render audio through OfflineAudioContext and check the samples. Pink noise measures -3.05 dB/octave against a nominal -3. MIT licensed. Happy to answer anything about the DSP.

SonicForge

A 16-channel tone generator with a live 3D spectrogram

A 16-channel tone generator with a live 3D spectrogram