Sound to MIDI Converter Online
Turn any sound — a recording, a melody, a field clip — into editable MIDI notes you can preview, fine-tune, and download in your browser.
- Any recording
- Piano roll preview
- Sheet music export
Click a note to hear it
How it works
Upload your sound
Drop an MP3 or WAV of any recording — a captured melody, an instrument, a sound clip. No account needed to convert and preview.
Preview and fine-tune
Hear the transcription on a piano roll, compare it to the original sound, and adjust detection sensitivity or minimum note length to re-transcribe.
Download MIDI or sheet music
Download the MIDI, edit notes on the roll, or export sheet music as MusicXML/PDF for MuseScore, Sibelius, or printing.
Formats & limits
- Input formats
- Sound
- Output
- MIDI
- File limits
- 50 MB · 10 min
- Price
- Free · no sign-up
Getting the cleanest result from a sound
- Pitched, tonal sounds transcribe well; unpitched noise, percussion, or ambient textures have no clear notes to detect.
- A single clear tone or melody line gives the most accurate MIDI — layered or noisy sounds add stray notes.
- Trim silence and background noise before uploading so detection focuses on the sound you care about.
- Use the built-in editor to clean up the transcription instead of re-recording.
Frequently asked questions
What kinds of sound can I convert to MIDI?
Any pitched audio, broadly — a hummed melody, an instrument, a recorded tone, a field recording with a clear tonal center, or a sound effect that has an identifiable pitch to it. The transcriber works by detecting note pitch and timing, so its output is only meaningful for sounds that actually have a pitch in the first place; clearly-pitched, sustained, or melodic sounds transcribe well, while anything without a stable fundamental frequency produces unreliable results regardless of what it sounds like to a human ear. If you're unsure whether your source qualifies, the simplest test is whether you could hum it back as a tune — if yes, it will likely transcribe reasonably; if it's just texture or noise with no singable line, don't expect usable notes back.
What formats are supported?
MP3 and WAV uploads are accepted, capped at 50 MB and roughly 10 minutes per file, which covers virtually any short recording or clip without needing to trim it first. Cleaner, less noisy recordings give more accurate transcriptions regardless of which of the two formats you use, since the model analyzes the decoded audio signal itself rather than anything specific to the file format — a high-bitrate MP3 and an uncompressed WAV of the same clean recording transcribe essentially identically. If your source is a longer field recording or ambient capture, isolating just the pitched section you actually want transcribed, rather than uploading the full recording, usually produces a cleaner and faster result than uploading the whole thing and discarding most of what comes back.
Is the sound to MIDI converter free?
Yes — converting and previewing is completely free, with no account required at any point. Upload an MP3 or WAV, watch the transcription build on the piano roll, and listen back against your original recording immediately to judge whether the result captured what you intended. What requires signing in is everything past that first pass: saving results to a persistent conversion history, using the full note-editing toolkit instead of just previewing, and exporting the corrected result as MusicXML or PDF sheet music alongside the plain MIDI download. A quick one-off MIDI download without any editing needs no account at all — the transcription itself is identical whether or not you sign in.
How accurate is it?
Transcription runs on Spotify's open-source Basic Pitch model, and its published research shows accuracy that varies by source material rather than being one fixed number — see our full benchmark breakdown by instrument for the actual figures. As a general rule, clean, single-line pitched sounds come out close to accurate with only occasional wrong notes, while noisy, unpitched, or layered audio produces meaningfully more errors that need manual correction on the piano roll. Treat any first transcription as a draft regardless of source type: delete stray notes, fix octave errors, and re-transcribe with adjusted sensitivity if too much is missing or extra — most sources improve noticeably after one or two adjustment passes.
Can it turn a clap, drum hit, or other unpitched sound into notes?
Not meaningfully. Basic Pitch detects pitch — its whole job is estimating which musical note is sounding and precisely when, frame by frame. A clap, a drum hit, footsteps, or white noise has no clear fundamental frequency for it to lock onto, so the model either detects nothing usable or produces a stray, unreliable note rather than anything resembling a real rhythm transcription. It's built specifically for tonal sounds — a hum, a sustained tone, a melodic instrument — not percussion or noise, so don't expect this to work as a substitute for dedicated drum-transcription or beat-detection software, which solve a genuinely different problem using different techniques entirely and don't share any of this pipeline's pitch-detection logic.
How is this different from Voice to MIDI or Audio to MIDI?
Same underlying transcription engine and the same Basic Pitch model doing the actual work — the three pages just frame identical functionality around different starting material, so picking the 'wrong' one costs you nothing in quality. Voice to MIDI is tuned in its framing specifically for singing and vocal melodies, Audio to MIDI is the general entry point covering any MP3 or WAV recording without a narrower focus, and this page covers anything pitched that falls outside those two more specific framings: a hummed idea, an instrument, a field-recorded tone, an ambient sample with a clear pitch. Use whichever title best matches what you actually have on hand — the conversion, editing tools, and export options behind all three are identical, so there's no accuracy penalty for picking the more general title.