Audio-to-MIDI Conversion

Converting audio to a symbolic MIDI representation makes the musical content editable. Pitch, rhythm, harmony and structure become data you can work with directly — rearranged, revoiced, quantised or analysed without further audio processing.

• Real-Time Processing
• File Conversion
• Conversion Options
• Workflows

Real-Time Processing

Click the green robot button to activate real-time conversion. The button turns magenta when active and a clock begins counting elapsed time. The algorithm detects pitch in the live audio stream and sends MIDI events to your DAW via the selected output and channel.

The Harmonics parameter is inaccessible during operation. Click the button again to deactivate before adjusting it.

In addition to MIDI note events, Transkr tracks and displays:

• Level — measured in dB RMS
• Tempo — measured in BPM
• Key — displayed as the detected key with its relative counterpart (e.g. Am = C means A natural minor, relative to C major)
• Detected notes — adapts to the number of simultaneous notes: a single note is shown by name in scientific pitch notation; two notes as an interval name; three or more notes as a chord symbol preceded by the Roman numeral of the scale degree relative to the detected key. When the detected bass note differs from the chord root, the chord is written as a slash chord (e.g. C/E).

File Conversion

To convert an existing audio recording, drag a WAVE, AIFF or MP3 file onto the green robot button. This initiates a new conversion and displays progress. The algorithm uses the current model settings during conversion.

On completion, a Standard MIDI File (SMF) appears in the same directory as the source audio. A text file is created alongside it containing tempo and other metadata — useful when importing into DAWs that require explicit tempo information, such as Pro Tools.

File conversion can also produce MusicXML or MXL (compressed MusicXML). The resulting score contains separate staves for Notes (treble and bass), Guitar (TAB), Melody, Guitar/Melody (TAB), and Chords. It can include chord symbols, including slash chords, chord diagrams, a tempo marking and a key signature. Chord formatting follows the chord settings described below.

File conversion uses forward-backward processing, which typically produces more accurate results than real-time forward-only processing. For file conversion, the standalone application is often more practical — it runs independently, without a DAW or processing chain.

The output MIDI file contains separate tracks for:

• Time signature
• Tempo
• Key signature
• Beats
• Notes
• Melody
• Chords

To select which tracks to include, right-click the green robot button and check the desired tracks in the popup menu before initiating conversion. The same menu also lets you set a default time signature, tempo and key signature.

Conversion Options

Right-clicking the green robot button opens a popup menu. Tracks with a checkmark are included in the output file; uncheck to exclude them. The tracks listed below have additional settings in submenus. Where available, Quantise aligns events to the selected quantisation level; the search direction determines whether the nearest beat is sought before or after each event.

Beats

Pulse

Detects the tactus, the main rhythmic pulse listeners tap to. This establishes a general rhythmic grid and the tempo in BPM.

Metre

Builds on the pulse by fitting duplets and triplets into the beat pattern to determine the full metric structure. Once a metre is detected, each beat position is weighted from weak to strong. Metre detection operates across five levels: semiquaver (sixteenth note), quaver (eighth note), crotchet (quarter note), minim (half note) and semibreve (whole note). Fastest and Slowest level set the upper and lower bounds of the metric range used in the analysis.

Key

Quantise

Aligns key signature change events to the selected level and search direction.

Melody

Texture

Sets whether the melody track is monophonic (a single melodic line) or biphonic (melody with bassline accompaniment). In biphonic mode, quantisation level and search direction can be set independently for the melody and the bassline.

Limit note length

Caps note duration to a maximum of one semibreve (whole note).

Quantise

Aligns note timestamps to the selected level and search direction.

Chords

Close voicing

Wraps chord notes into the selected octave register rather than stacking upward from the root. This centres voicings around a consistent register and produces smoother voice leading as the harmony changes. When disabled, chords are built in root position.

Slash chords

When the detected bass note differs from the chord root, includes it in the chord symbol (e.g. a C major chord over a detected E in the bass is written as C/E). When disabled, chords are written by root and quality alone.

Simplify chords

Reduces detected chords to a basic triad or dominant seventh: major and augmented chords become a major triad, minor, diminished and half-diminished chords become a minor triad, and dominant chords become a dominant seventh. Other chord types are left unchanged.

Omit notes

Removes chord factors following standard voicing conventions: the fifth is omitted from extended chords, and additional tones are omitted from dominant seventh, ninth, eleventh and thirteenth voicings according to standard practice.

Limit chord length

Caps chord duration to a maximum of one semibreve (whole note).

Quantise

Aligns chord timestamps to the selected level and search direction.

Model

Selects the harmony-analysis model: Generative or Discriminative.

Chord names

Accepts a comma-separated list of chord symbols to match and time-align with the audio. Examples: Cmaj,Gmaj,Amin,Fmaj. For songs, using a stem splitter to isolate the harmonic part first improves accuracy. See supported chord symbols.

Octave

Sets the register for chord output using Helmholtz notation, from Sub contra to 6 line. The default is 1 line (middle C). When Close voicing is enabled, this register is the centre around which chord notes are wrapped.

Remove silence

Strips silence from the beginning and end of the audio stream before conversion. Disable this if you want the MIDI data to remain temporally aligned with the original audio recording.

Align level

Measures the input file's signal level and applies gain to bring it to a target reference level before conversion, so files recorded at different levels start from the same level before detection.

Workflows

Recomposition and revoicing

A MIDI representation of a recording makes the harmonic and melodic content fully editable. Individual notes can be repositioned, chords revoiced, melodies transposed and rhythmic values adjusted — all without re-recording. This is particularly useful when working from a sketch or improvisation where the musical idea is clear but the arrangement needs development.

Working with generative audio

Generative AI audio tools produce output that can be convincing in texture but difficult to direct at the level of individual notes or chords. Converting generated audio to MIDI provides a symbolic layer that can be edited, corrected and used to drive different sound sources — restoring intentional control over the musical content. Tools that convert MIDI back to human-like performance audio can extend this workflow further.