WAV TO MIDI · FREE · RUNS IN YOUR BROWSER · UPDATED 2026-11-07
WAV to MIDI Converter
You can convert a WAV file to MIDI with the same converter used for MP3. Choose your .wav file on the homepage, convert it in your browser and download a standard .mid file with the detected notes. Files up to 50 MB and 3 minutes are accepted. Nothing is uploaded to a server.
Updated 2026-11-07 · WAV preflight tables (onset risk, channel pick, clip container, note-density gate).
How to convert, step by step
- Open the converter. Go to the MP3toMIDIFlow homepage. The converter is at the top of the page; nothing needs to be installed.
- Choose your file. Select or drag in one WAV file. Standard PCM WAV under 50 MB and 3 minutes works best.
- Convert in the browser. Select Convert to MIDI and keep the tab open. The note-detection model and decoder load from this site and run on your device; your audio is not uploaded.
- Download and edit. Download the .mid file and open it in a DAW or notation editor to set the tempo, fix wrong notes and choose an instrument sound.
Why convert WAV rather than MP3?
WAV is usually uncompressed PCM audio, so it keeps everything the recording captured. MP3 throws away detail that the encoder judges inaudible. For note detection the difference is often small with a clean, well-encoded file, but WAV removes one possible source of error, so if you recorded the audio yourself, export WAV from your recorder or DAW and convert that. We ran a same-melody WAV vs MP3 test on one short melody; read it for what it shows and what it does not.
A practical point about size: uncompressed stereo WAV at 44.1 kHz and 16 bits uses about 10 MB per minute, so a full three-minute stereo WAV is roughly 30 MB — under the 50 MB limit. 24-bit or high-sample-rate WAV files are larger; at 96 kHz/24-bit stereo, three minutes is well over 50 MB. If your file is too big, export a 44.1 or 48 kHz version or a shorter section. The converter resamples everything to 22,050 Hz mono for analysis anyway, so a higher sample rate does not improve the result.
How the WAV is turned into notes
The browser decodes your WAV (or, if the browser's own decoder cannot read it, a bundled WebAssembly FFmpeg decoder does). The audio is mixed to mono and resampled, then analysed by Basic Pitch, an open-source note-detection model from Spotify. Basic Pitch estimates which pitches are sounding over time and where notes begin. Those estimates are turned into note events — pitch, start time, length and a velocity from the note's loudness — and written into a single-track MIDI file.
The model is not tied to one instrument and can detect more than one note at a time, so a WAV of a piano chord or a strummed guitar can produce several simultaneous notes. That is different from separating instruments: everything audible ends up on the same track.
Tips for a better WAV to MIDI result
- Record or export one instrument. If you made the music in a DAW, solo the part you want and bounce it to WAV instead of converting the full mix.
- Keep it dry. Reverb and delay tails are heard as extra, lingering pitches. Bounce without effects where you can.
- Avoid clipping and heavy compression. A healthy level with peaks below 0 dBFS is fine; very quiet files may be rejected as silent.
- Trim leading silence and noise. Hum, hiss and room noise can create short false notes.
- Convert sections. For a long performance, split it into parts under three minutes and join the MIDI clips in your DAW.
What WAV to MIDI cannot do
Converting audio to MIDI is transcription, not file conversion. The MIDI file contains notes, not the sound of your WAV, and it will sound like whatever instrument your software assigns. Drums and percussion are not transcribed as a drum part: the model looks for pitched notes, so drum hits may disappear or show up as stray pitched notes. Full songs with vocals, bass, drums and chords produce a dense, overlapping result that needs heavy editing. No automatic transcription is note-perfect, so plan to listen back and correct.
Supported WAV variants are those your browser or the bundled FFmpeg decoder can read; standard PCM WAV works. Damaged files or unusual codecs inside a WAV container may fail to decode, and the converter will tell you rather than produce a made-up result.
Limits that apply to every file
- One file at a time, up to 50 MB and 3 minutes. Trim longer recordings into sections first.
- Mono analysis. The audio is mixed down to one channel at 22,050 Hz before note detection. If the left and right channels cancel each other out, see why a file that plays can be reported as silent.
- One instrument track, fixed tempo. All detected notes go on a single MIDI track. The file is written at 120 BPM; the converter does not detect the song's tempo, key or time signature, so bar lines in your DAW will not line up with the music until you adjust them.
- No pitch-bend data. Each detected note is written as a plain note-on/note-off with a velocity based on its loudness.
Bit-depth × sample-rate → onset-risk checklist (WAV → MIDI)
Conclusion first: for this site’s converter, every WAV is resampled to 22,050 Hz mono before note detection, so 24-bit/96 kHz does not buy cleaner onsets. Use the checklist below to pick a safe export instead of chasing higher container specs.
This is a table molecule (not a FAQ restatement). Rows are WAV-container choices; the risk column is about false/missed onsets after the forced resample, not studio mastering quality.
| WAV export | Approx size / min (stereo PCM) | What the model actually hears | Onset-risk band | Action before convert |
|---|---|---|---|---|
| 16-bit · 44.1 kHz | ~10.1 MB/min | Downsampled to 22.05 kHz mono | Low — default sweet spot | Keep; trim silence; stay under 50 MB / 3 min |
| 16-bit · 48 kHz | ~11.0 MB/min | Same 22.05 kHz mono path | Low | Keep; no need to re-render to 44.1 first |
| 24-bit · 48 kHz | ~16.5 MB/min | Bit depth discarded after decode/resample | Low–medium — size risk, not pitch gain | Re-export 16-bit if the file nears the 50 MB cap |
| 24-bit · 96 kHz | ~33.1 MB/min | Heavy downsample; HF content unused for pitch | Medium–high — often fails the size gate first | Bounce 44.1/48 kHz 16-bit; do not expect better MIDI from 96 kHz |
| 32-bit float · 48 kHz | ~22.1 MB/min (IEEE float) | Decoded then resampled; float headroom ≠ more notes | Medium — codec/browser variance | Prefer PCM 16-bit integer WAV if decode fails |
Interactive onset-risk picker
Pick the bit depth and sample rate of the WAV you plan to send. The readout maps to the checklist above for this converter only.
Result: choose values, then check.
Stopping rule: if the export is 24-bit/96 kHz only to “help MIDI,” stop — re-bounce at 44.1 or 48 kHz 16-bit. Onset quality here is gated by dryness, single-stem clarity, and level — not by unused ultrasonic bandwidth.
Mono fold / channel-pick rule table (stereo WAV → pitch-track channel)
Conclusion first: this converter folds stereo to one channel before analysis. If left and right disagree (hard pan, out-of-phase doubles, mid/side tricks), the fold can cancel the melody or bury it. Pick which channel becomes the pitch track before you bounce.
| Stereo layout you have | What mono-fold does | Channel to bounce | Skip / warning |
|---|---|---|---|
| Identical L = R (dual mono) | Fold is harmless (same signal twice) | Either; dual-mono WAV is fine | None for channel pick |
| Melody hard-panned Left, pad Right | Fold mixes pad into the pitch track | Export Left only (mono bounce of L) | Do not send the full stereo mix if the pad is loud |
| Melody hard-panned Right | Same risk mirrored | Export Right only | Verify with headphones which side carries the line |
| Out-of-phase / inverted double | Fold can near-cancel the fundamental | Pick the louder coherent side; or sum with polarity fix first | If the file “plays” but converts as silent, see the stereo-cancellation guide |
| Mid/Side or width plugin left engaged | Side energy becomes noise-like onsets after fold | Bounce a mono Mid (or collapse width to 0%) then export | Avoid “extra wide” masters for transcription |
| Call/response on opposite sides | Fold stacks two melodies into one track | Split: bounce L phrase, then R phrase, as two WAVs | One stereo file → one tangled MIDI track |
Quick channel-pick helper
Pick: choose a layout.
Method note: channel-pick is a pre-bounce rule for pitch tracking on a mono-fold converter. It is not stem separation and not a drum-map tool.
Clip / true-peak preflight mini-table (WAV-container scoped)
Conclusion first: this preflight is about the WAV container you will upload — bit-depth integer ceilings, float peaks, and whether the file’s duration×rate still fits the 50 MB / 3 min gate. It is not the hub’s peak-headroom checklist.
| Container check | How to read it on a WAV | Pass for this converter | Hold / fix |
|---|---|---|---|
| Integer peak vs 0 dBFS full-scale | 16-bit max sample ≈ ±32767; 24-bit ≈ ±8.3M. Meters at 0 dBFS mean the integer ceiling is hit. | Peaks below 0 dBFS on the region you will send; no brick-wall samples on the attack you care about. | If the WAV already hits the integer ceiling, re-export from an unclipped source — lowering gain on a clipped integer WAV does not restore peaks. |
| 32-bit float peaks > 1.0 | Float WAV can store samples beyond ±1.0. Players may show >0 dBFS without “clip” flags until rendered to integer. | Float peaks normalized so render-to-16-bit stays under 0 dBFS, or bounce integer 16-bit with headroom first. | Do not assume “float = unclipped forever” after a later integer bounce. |
| Header duration vs wall-clock | WAV data size ÷ (sample rate × channels × bytes/sample) should match the play length you expect. | Declared length ≤ 3 minutes and matches what you hear. | Truncated or junk trailing chunks: re-export a clean PCM WAV of the phrase only. |
| Size honesty (MB vs format) | Stereo 16/44.1 ≈ 10 MB/min; 24/96 ≈ 33 MB/min. A “3-minute” 24/96 stereo file often exceeds 50 MB. | File ≤ 50 MB after the format you chose. | If over cap, change bit-depth/SR or shorten the clip — do not compress to MP3 just to sneak size (that reintroduces encode loss). |
| Codec-in-WAV oddities | Some “.wav” files wrap ADPCM/mp3-in-wav. Browsers or the bundled decoder may reject them. | Standard PCM (or float) WAV decodes without error. | Re-wrap/bounce to PCM WAV; a failed decode is preferred to a fake MIDI. |
Stopping rule: if the WAV header says you are over 50 MB or over 3 minutes, fix the container (shorter / leaner PCM) before opening the converter. This block stays WAV-scoped on purpose — it does not reuse the hub peak/normalize checklist.
WAV size → expected MIDI note-density honesty gate
Conclusion first: a large WAV does not mean a dense, useful MIDI file. Size mostly tracks duration × sample rate × bit depth × channels. Note density follows how many pitched onsets the model can lock — a 40 MB noisy mix can yield fewer usable notes than a 4 MB dry mono line.
| WAV profile | Typical size cue | Expected MIDI note density | Honesty gate |
|---|---|---|---|
| Dry mono melody, ≤60 s, 16/44.1 | ~2–6 MB | Sparse–moderate: one line, clear onsets | Pass — best case for editing |
| Stereo piano chords, 1–2 min, 16/48 | ~10–25 MB | Moderate–dense polyphony on one track | Pass with edit plan — expect overlapping notes to clean |
| Full mix (vocals+band), 2–3 min | ~20–50 MB | Very dense / tangled; many false overlaps | Hold — bounce a stem first; size ≠ usable notes |
| 24/96 stereo “hi-res” of a short riff | Often 15–40 MB for <90 s | Same as a lean 16/44 bounce of that riff | Re-bounce — you are paying MB for unused bandwidth |
| Near-silent / noise floor only | Any size | Near-zero usable notes (or spam micro-notes) | Fail — raise a clean take; do not “gain up” noise |
Size → density estimator
Enter approximate WAV size (MB) and duration (seconds). The gate is an honesty check for this converter’s 50 MB / 180 s caps — not a prediction of note-perfect transcription.
Gate: enter size and duration.
Sources / method: size cues use standard PCM byte-rate maths (sample rate × channels × bytes/sample). Cap numbers (50 MB, 3 minutes, 22,050 Hz mono analysis) match this site’s published converter limits. Note-density bands are honesty guidance from how Basic Pitch–style pitched onset models behave on dry stems vs mixes — not a measured accuracy %.
Questions
Is the WAV to MIDI converter free?
Yes. There is no account, no payment and no daily quota. Each file can be up to 50 MB and 3 minutes long.
Is my WAV file uploaded?
No. Decoding and note detection run in your browser. The site only delivers the converter, model and decoder files.
Does a 24-bit or 96 kHz WAV give better MIDI?
Not in this tool. Audio is resampled to 22,050 Hz mono before analysis, so a 44.1 or 48 kHz export is enough and keeps the file under the size limit.
Can it convert a WAV of drums to MIDI?
Not reliably. The model detects pitched notes; it does not produce a drum map, so percussion is mostly missed or turned into stray notes.
Need a single browser tool that also accepts MP3/M4A? Use the audio to MIDI converter on this site, or browse the audio tools hub.