How does WSOLA change audio speed without shifting pitch?
Waveform Similarity Overlap-Add (WSOLA) decouples playback duration from vocal pitch by splicing overlapping 40 ms audio grains rather than changing the playback sample rate. Searching within a local +/- 10 ms tolerance window aligns adjacent waveform phases before blending them, keeping voices natural across 0.5x to 2.5x speeds.
Splicing 40 ms Hann windows with overlap-add correlation
After your browser decodes compressed tracks under the W3C Web Audio API 1.1 Specification, the worker receives planar Float32Array channels. WSOLA builds a periodic Hann window around 40 ms long (50% overlap hop) and normalizes each sample against the summed window curve.
At every synthesis step, the engine evaluates candidate offsets within +/- 10 ms around the target timeline position. Picking the peak normalized cross-correlation against the prior grain keeps harmonic periods aligned, avoiding hollow phase cancellations. If your source lecture still contains long gaps, strip awkward pauses from lectures beforehand for tighter pacing.
Unchecking pitch lock for classic varispeed resampling
Clearing Preserve pitch independently when changing speed (WSOLA time-stretch) bypasses overlap-add splicing entirely. Instead, the worker multiplies your chosen Playback speed by the semitone ratio and runs single-pass cubic Catmull-Rom interpolation across every channel.
Varispeed coupling behaves exactly like speeding up or slowing down analog tape. Raising speed to 2.0x halves duration and raises pitch by one full octave (+12 semitones), while eliminating overlap-add grain boundaries on percussive synth loops or drum stems.
Transposing musical keys from -12 to +12 semitones
Musicians practicing difficult guitar solos or vocalists adapting karaoke backing tracks often need exact half-step key shifts without changing song BPM. Moving Pitch shift (semitones) scales musical frequency by 2^(semitones / 12) across one octave downward (-12 st) to one octave upward (+12 st).
Combining cubic Catmull-Rom resampling with time-stretching
Because channel arrays read from an MDN Web Audio AudioBuffer interface contain raw floating-point PCM amplitudes, shifting key at constant tempo uses a clean two-step pipeline. First, WSOLA stretches or compresses the clip by pitchRatio / speed so its intermediate length compensates for resampling.
Second, 4-point Catmull-Rom cubic interpolation resamples the stretched buffer by pitchRatio, shifting pitch smoothly while landing on the exact target duration. Before transposing a full rehearsal track, you can also extract just the chorus loop using our Audio Cutter & Trimmer.
Using quick speed presets for study and transcription
Six quick preset buttons (0.75x, 0.9x, 1.0x, 1.25x, 1.5x, and 2.0x) sit right beside the track header for rapid auditioning. Clicking any preset updates Playback speed immediately and re-renders the 160-bar Processed audio waveform inside everyday.worker.ts.
Dragging Playback speed adjusts your speed multiplier in 0.05x steps from 0.50x up to 2.50x, while clicking Apply speed & pitch commits slider adjustments together. Output filenames automatically embed your parameters, such as -1.25x-minus2st.wav, so multiple practice versions never overwrite each other.
Recommended speed, semitone, and pitch-lock settings
Choosing independent pitch lock versus coupled varispeed depends on your audio goal, such as accelerating spoken narration, transcribing fast solos, or shifting vocal ranges. Standard mobile playback containers like those described in the Library of Congress MPEG-4 AAC format guide preserve these rendered tempos across phones and tablets.
| Audio task | Playback speed | Pitch shift | WSOLA pitch lock | Practical result |
|---|---|---|---|---|
| Podcast or lecture review | 1.25x to 1.75x |
0 st |
Checked (On) | Cuts listening time without chipmunk voices |
| Language or dictation drill | 0.75x to 0.90x |
0 st |
Checked (On) | Slows rapid consonants while keeping natural tone |
| Guitar or vocal key transpose | 1.00x |
-4 st to +4 st |
Checked (On) | Changes musical key while locking original BPM |
| Complex music transcription | 0.75x |
0 st |
Checked (On) | Reveals fast fingerings at original tuning |
| Lo-fi slowed or sped effect | 0.85x or 1.15x |
0 st |
Unchecked (Off) | Shifts pitch and tempo together like analog tape |
Musicians matching an instrument tuning reference after transposing a backing track can pair exported clips with reference musical pitch tuning tones in the same session. Always audition short percussive sections when pushing speed past 2.0x so drum transients remain crisp.
Tip: Click Load sample voice clip to audition
sample-voice.wavat the default1.25xspeed first, then compare checking and unchecking Preserve pitch independently when changing speed (WSOLA time-stretch) at1.5xto hear phase alignment firsthand.
Exporting permanent WAV, M4A, and WebM audio offline
Unlike browser extensions that only alter playback speed temporarily inside an HTML media element, every render here generates fresh PCM samples. Downloading exports an independent audio file that opens at your new tempo and key in any desktop editor or phone player.
Encoding 16-bit WAV alongside WebCodecs AAC and Opus
Clicking Download .wav packages lossless 16-bit RIFF PCM samples directly from the worker buffer at your source sample rate (4,000 Hz to 192,000 Hz). In browsers supporting WebCodecs AudioEncoder, Download .m4a (128 kbps AAC) and Download .webm buttons also appear.
WebCodecs encodes WebM streams using the IETF RFC 6716 Opus audio codec, automatically resampling PCM audio to 48,000 Hz prior to muxing. Choose WAV to export uncompressed masters for studio DAWs, or pick M4A and WebM to keep mobile downloads compact.
Processing private rehearsals and lectures in client RAM
Confidential client interviews, unpublished band demos, and recorded university seminars remain strictly inside browser RAM from drop to download. Because everyday.worker.ts runs all WSOLA correlation loops off the UI thread, long clips process smoothly without stalling page scrolling.
Students pairing dense reading assignments with custom-paced spoken recordings can also listen to document narrations offline at custom speed or bring recorded lecture audio straight into this tab. When you close the browser tab, object URLs revoke automatically and leave zero temporary copies behind.