Voice Changer Online - Pitch, Robot & Echo Effects

Every effect on this page runs offline in your browser and is written out as a standard 16-bit WAV file. Because the processing is deterministic, the same input with the same settings always produces the same output, and a typical clip renders in a second or two. The pitch effects preserve duration — raising or lowering the pitch does NOT change how long the clip is, so a five-second recording stays five seconds whether you shift it up an octave or down. Under the hood the pitch and speed processing runs on Signalsmith Stretch, an open-source, studio-grade time-stretch and pitch-shift engine compiled to WebAssembly and driven entirely offline in your browser: it analyzes the audio spectrally in overlapping blocks and rebuilds it at the new pitch and tempo, which keeps voices natural and transients tight even at large shifts. If the WASM engine cannot load — an old browser, or WebAssembly blocked by policy — the tool automatically falls back to a built-in time-domain shifter called WSOLA, short for waveform-similarity overlap-add: the audio is chopped into overlapping grains of about 90 milliseconds, the grains are re-spaced in time, and at every splice the next grain is slid back and forth by up to 10 milliseconds to the position where its waveform lines up best with what came before, then a final resample lands the duration exactly back where it started. Either way every frequency moves by the factor shown in the readout — 2 raised to the power of the semitone shift divided by 12 — and after processing the result panel states which engine produced the render: Signalsmith (studio quality) or WSOLA (fallback). Speed is a separate, independent control. Choosing 0.5× or 2× runs the same engine as a pure time stretch, so the clip gets longer or shorter while the pitch stays put. You can combine both: a +5 semitone chipmunk at 1.5× speed is a single pass through the engine. If you miss the old-school sound, the Tape mode toggle switches to classic playback-rate resampling, where pitch and speed are coupled exactly like speeding up a tape or spinning a record at the wrong RPM — +12 semitones then plays at double speed and half duration, and the readout shows the duration factor so there are no surprises. The robot effect is ring modulation: your voice is multiplied by a 30 Hz sine wave, replacing the natural harmonics with sum and difference frequencies — the same technique behind the 1960s Dalek voices. The echo effect feeds the signal through a 250 millisecond delay with 45 percent feedback and adds a second and a half of tail so the last repeats ring out. Practical uses go well beyond pranks: podcasters anonymize interviewee voices with a few semitones of shift while keeping natural pacing, game developers batch-generate creature voices, teachers demonstrate how pitch and tempo relate, and producers use tape mode as a lo-fi texture. Because granular processing is CPU-intensive, input is capped at 10 minutes; everything runs locally, so sensitive recordings never touch a server.