Pitch & notes
Frequency tracking converts confident monophonic pitch regions into MIDI note candidates, with editable timing and pitch.
Turn recorded melodies into a playable pitch map, editable note list and downloadable MIDI. All locally, with clear confidence and limitations.
Listen, analyze, refine and export without uploading audio.
Choose a vocal, instrument, or solo melody recording. Clear, isolated melody audio works best.
MP3, WAV, M4A, OGG, FLAC or other browser-decodable audio · Max 180 seconds · Files stay on your device.Find useful musical clues while keeping the final corrections in your hands.
Frequency tracking converts confident monophonic pitch regions into MIDI note candidates, with editable timing and pitch.
Amplitude-onset patterns estimate beats and possible changes in activity. Uncertain results stay explicitly uncertain.
Export MIDI for a DAW or download CSV/JSON for further review. Keep your original recording untouched.
No. This first version estimates dominant pitch from mixed audio. It does not isolate vocals, instruments or polyphonic notes. A clean solo recording gives the most useful results.
No. This version uses free, self-contained signal-processing algorithms in JavaScript. “Melody Extractor” is the tool name; no neural model or paid API is required.
MIDI stores note events and timing, not audio timbre. Pitch and tempo estimates are imperfect; manually edit detections and choose a sound in your music software.
No. Your audio and analysis remain in your browser. You must choose to download exported files.
Import a supported recording or generated demo, choose analysis settings and estimate the melody. Review and edit notes against the original sound before export. MIDI contains note events, not the source audio.
This is a conventional monophonic pitch estimator, not neural transcription, source separation or reliable polyphonic analysis. Tempo/sections/confidence are estimates; dense music, reverb and octave ambiguity can mislead. Semitone MIDI omits pitch bend/microtonality. Same-pitch overlaps are rejected rather than silently clipped. Whole-file decoding, file/duration/channel/memory budgets and browser codecs apply.
Import supported audio or use the generated demo, run local pitch analysis, and review the estimated note contour against the original recording. Correct note pitches and timing, then export MIDI events, CSV or JSON.
Authored example: Analyze a simple single-note keyboard phrase with little reverb. Compare the estimated contour with each original note and correct an octave error before exporting.
Limits: This slug retains an AI label, but the current extractor uses conventional YIN-style pitch and onset heuristics without neural models. It estimates one dominant melody; it does not isolate vocals/instruments or transcribe polyphonic arrangements reliably. Tempo, sections and confidence are estimates. MIDI stores note events rather than audio and omits pitch bend/microtonality; same-pitch overlaps are rejected.
No. It estimates a dominant monophonic line and does not separate instruments or reliably transcribe polyphony.
No. MIDI contains editable note events, not the source audio or its timbre.