Audio to MIDI Converter

MIDI is a performance format — it captures notes, timing, and velocity but not visual notation. MusicXML is a notation format that describes how to render sheet music (beams, slurs, key signatures). This tool exports MIDI. If you need MusicXML, open the .mid in MuseScore (free) and export MusicXML from there.

Basic-pitch (Spotify Research, 2022) is open-source, runs on CPU, and handles both monophonic and polyphonic audio in one model. Onsets-and-Frames is piano-specific. MT3 covers more instruments but is slower and needs a GPU. For a fast, low-cost converter, basic-pitch is the right tradeoff.

Yes — basic-pitch detects polyphony, so chords arrive as simultaneous note-on events in the output MIDI. You can split them into individual notes or treat them as a chord block in your DAW. For best chord accuracy, use stems rather than full mixes.

Drag the .mid file from Finder or Explorer into Ableton's Session view. It creates a new MIDI track with all notes preserved. Assign an instrument plugin (Operator, Wavetable, or any VST) to play the MIDI.

File → Open and select the .mid. Logic creates a project with each MIDI channel as a track. Alternatively, drag the .mid into an existing project's tracks area.

Inference runs at roughly 2x real-time on a CPU worker — a 10-second clip takes about 4-5 seconds of model time, plus upload and queue wait. Longer files scale roughly linearly.

Yes. Both accept .mid files directly. Expect to fix the rhythm: the model returns exact note timings, not a quantized beat grid, so bar lines and note values usually need cleanup in the notation editor.

Your audio is uploaded to our inference provider to run the model and is not kept by us. The resulting .mid is copied to our own storage so it stays in your Library and you can re-download it later.