Refentra
English
Go to tools

Change pitch at constant duration

±12semitones, STFT phase-vocoder, identical frame count/duration. Formants, transients and stereo phase are approximate; frequencies above Nyquist are omitted.

  • Runs in your browser
  • No upload
  • Check before download

Audio stays on this device. No upload or microphone recording. ±12semitones, STFT phase-vocoder, identical frame count/duration. Formants, transients and stereo phase are approximate; frequencies above Nyquist are omitted. Files up to 16 MiB, mono or stereo at 32, 44.1 or 48 kHz, up to 60 seconds and 24 MiB of PCM; the 20-second deadline includes reading and native decoding. DSP exports 16-bit WAV, drops source metadata and warns before clipping. The original bytes stay available. MP3 ID3 edits: up to 120 seconds, ID3v2.3 or 2.4, tags up to 2 MiB, PNG or JPEG cover up to 1 MiB and 4 MP; the audio frames are unchanged. Unsupported, encrypted or signed data is rejected. Supported formats: WAV (PCM 16, 24 or 32 bit, or float32 with |x| up to 16), MP3 (MPEG-III, ID3v2.3 or 2.4) and M4A (AAC-LC).

Supported model and limits

How to use this tool

  1. Choose audio

    Load a supported local audio file and choose a pitch shift from −12 to +12 semitones. Preview the processed sound before you export a WAV.

  2. Supported model and limits

    A 440 Hz test tone shifted +12 semitones becomes approximately 880 Hz. The frame count, sample rate and duration stay the same, and the export is a 16-bit WAV.

  3. Download

    Compare the saved sound with the original, listening for artifacts and checking for clipping warnings. A correct test-tone frequency doesn't promise a natural-sounding voice.

Supported model and limits

What does this result mean?

The STFT phase vocoder shifts the frequency while keeping the duration. Formants, transients and stereo phase are only approximate, and content above Nyquist is left out.

How do I resolve rejected or missing input?

Choose mono or stereo at 32, 44.1 or 48 kHz, with input of at most 60 s/16 MiB. Unsupported or encrypted files are rejected. Use a smaller input instead of ignoring decode errors.

What are the limits of this tool?

Audio stays on this device, with no microphone recording or upload. PCM ≤24 MiB, and the 20 s deadline includes reading and decoding. Source metadata is dropped, and your original remains available.

Some audio workflows use Mediabunny1.60.0 (MPL-2.0). Copyright notices and original license · Download preferred TypeScript source1.60.0