Make a whole new instrument with your voice!
Sing a note, teach a synth your tone, then play a melody with your voice-instrument ♡
Sing one steady note.
Hold your own steady sound for 2 seconds—a hum, a vowel, or an unusual tone. No perfect pitch needed.
Allow your microphone when asked. Your voice is tuned for the piano. Compare it with a learned synth.
Your recording stays in your browser.
A synth that learns from you.
These waveforms and the fit score describe the harmonic model only. They do not measure how much the synth sounds like your original recording.
Your voice. All these notes.
Keep your tone as the notes change.
LEARNED WITH HANDWIREDWhat did it learn?
Record or train above to make this keyboard sound like your chosen tone.
Computer key guide ⇧ = HOLD SHIFT
Shape the soundOPTIONAL · 64 HARMONICS ＋
Drag a slider to change the tone. Click its number to lock it during training.
First, find your sound. Rust measures the pitch and 64 harmonics in your recording—the ingredients that give a note its character.
Then, teach the synth. Handwired builds a little trainable program that mixes those harmonics into a wave. It measures the difference from your target and calculates how each harmonic should change. As it learns, the sound gets closer.
Finally, make a melody. Choose Voice to play a loop from your actual recording, tuned to each key. Choose Learned synth to play Handwired’s harmonic approximation. In Voice mode, Handwired also learns a different tone-correction filter for each nearby note. Your recording supplies the texture; the learned filters help keep its tonal color when the pitch changes.
The nerdy details ＋
Handwired’s Circuit combines 64 sine waves into 256 waveform samples, computes mean squared error against a phase-aligned target, and backpropagates that error. Gradient descent updates the 64 amplitudes; locked sliders stay fixed. Rust/WebAssembly handles pitch detection and harmonic analysis. Web Audio plays the result.
The model-fit percentage measures the simplified harmonic target. The learned synth models harmonic amplitudes; speech, breath, phase, and changes over time aren’t modeled. Voice mode loops recorded PCM with a short crossfade and changes its playback rate. A second Handwired Circuit learns 20 smooth filter controls per note within one octave above and below the tuned recording. Its loss compares smoothed spectral envelopes, with penalties for excessive or abrupt changes. Filters are rendered in a background worker, level-matched, and kept only if they improve the measured envelope on a separate part of the recording. They do not recover missing harmonics or preserve the speed of vibrato and breath fluctuations; this is experimental tone correction, not full formant-preserving synthesis. Outside the learned range, playback uses plain transposition. There’s no pretrained model, and no audio is uploaded.