Guide

MIDI Format: The Complete Technical Guide

PC By Pablo Cirre • Updated May 7, 2026

MIDI Format: The Complete Technical Guide

MIDI (Musical Instrument Digital Interface) is not an audio format — it contains no audio data. Instead, it is a protocol and file format that stores musical performance information as discrete events: which note was played, when, for how long, how hard, on which instrument channel, and with which controller changes (pitch bend, modulation, expression). A MIDI file is essentially a recipe for a performance that a synthesizer, sampler, or software instrument reads and plays back. Standardized in 1983 by the MIDI Manufacturers Association (MMA), MIDI transformed the music industry and remains foundational to music production 40 years later.

Historical Context

Before MIDI, synthesizers from different manufacturers could not communicate. Each had proprietary connections; connecting a Minimoog keyboard to a Roland synthesizer was impossible without custom wiring. MIDI was developed collaboratively by Sequential Circuits, Roland, Korg, Yamaha, and others as an open standard — revolutionary for the era. The first public MIDI demonstration was at NAMM 1983, connecting a Sequential Circuits Prophet-600 to a Roland JP-6. Dave Smith (Sequential Circuits) is widely credited as MIDI's primary inventor. MIDI's design for 31,250 baud serial transmission over 5-pin DIN connectors was chosen for cost: standard UART chips were cheap in 1983. The same protocol still ships hardware in 2024, though USB-MIDI and MIDI over IP (RTP-MIDI/AppleMIDI) are now standard.

MIDI 1.0 Physical Layer

Traditional MIDI uses:

5-pin DIN connectors: MIDI IN (receive), MIDI OUT (transmit), MIDI THRU (pass-through copy of IN)
31,250 baud serial, 8N1: 8 data bits, no parity, 1 stop bit. Each byte takes 320 µs to transmit.
Optoisolator on MIDI IN: Isolates ground loops between instruments
Maximum cable length: 15 meters (specified), though 50+ meters often works in practice

MIDI 2.0 (2020) upgrades to 32-bit messages and bidirectional protocol negotiation while maintaining backward compatibility.

MIDI File Format (SMF — Standard MIDI File)

MIDI files are Standard MIDI Files (SMF). They have two parts:

Header Chunk

4D 54 68 64  = "MThd" (magic)
00 00 00 06  = Chunk length: always 6 bytes
00 01        = Format: 0, 1, or 2
00 XX        = Number of tracks
E7 28        = Division (timing resolution)

Format types:

Format 0: Single track containing all channels. All events in one track. Simplest format; compatible with all hardware.
Format 1: Multiple synchronous tracks. First track = tempo map; additional tracks = individual instruments. Most common for multi-track sequencer files.
Format 2: Multiple independent patterns (less common; used for song libraries).

Division (timing):

Ticks Per Quarter Note (PPQN): e.g., 480 TPQN means 480 ticks per beat. Tempo (in the Set Tempo event) determines actual time per tick.
SMPTE timecode: Division encodes SMPTE frame rate + subframes per frame for video sync.

Track Chunk

Each track chunk: 4D 54 72 6B ("MTrk") + 4-byte length + MIDI events.

Events are preceded by a variable-length delta time (1–4 bytes, MSB-first with continuation bit):

0x00 = 0 ticks (simultaneous with previous event)
0x60 = 96 ticks
0x81 0x00 = 128 ticks (2-byte encoding)

MIDI Event Types

Channel Voice Messages

The core of MIDI performance data. Channel number is in the low nibble of the status byte.

Status Nibble	Event	Data Bytes	Description
0x8n	Note Off	note (0–127), velocity (0–127)	Release note on channel n
0x9n	Note On	note, velocity	Press note. Velocity=0 = Note Off
0xAn	Aftertouch	note, pressure	Per-note pressure (rare)
0xBn	Control Change (CC)	controller (0–127), value (0–127)	Modulation, volume, pan, etc.
0xCn	Program Change	program (0–127)	Select instrument/patch
0xDn	Channel Pressure	pressure (0–127)	Overall channel aftertouch
0xEn	Pitch Bend	LSB (0–127), MSB (0–127)	Pitch deviation ±2 semitones (default)

Note numbers: Middle C = note 60 (MIDI C4 convention). Range 0–127 = C-1 to G9.

Velocity: 0–127 represents note attack strength. Velocity 0 on Note On is equivalent to Note Off (running status optimization).

Important Control Change Numbers

CC#	Controller	Purpose
1	Modulation Wheel	Vibrato depth, LFO
7	Channel Volume	Overall channel level
10	Pan	Stereo position (0=left, 64=center, 127=right)
11	Expression	Relative volume (CC7 × CC11 / 127)
64	Sustain Pedal	Hold notes (0–63=off, 64–127=on)
91	Reverb Send	Effect return level
93	Chorus Send	Effect return level
120	All Sound Off	Immediately silence all notes
121	Reset All Controllers	Restore CC defaults
123	All Notes Off	Release all held notes

General MIDI (GM)

General MIDI standardizes Program Change assignments so that a GM-compatible synthesizer always responds to the same instrument on the same program number. Channel 10 is reserved for drums (note number = drum sound). Piano = 0, Electric Bass = 33, Violin = 40, Trumpet = 56, Acoustic Guitar = 25. 128 instruments + 47 drum sounds.

GM Level 2 (GM2) extends to 256 instruments, 16 drum kits, and additional controllers.

Meta Events

Meta events are SMF-specific (not transmitted over MIDI hardware). They carry non-performance data:

FF type	Name	Data
FF 01	Text Event	Arbitrary text
FF 02	Copyright Notice	Copyright string
FF 03	Track Name	Track/sequence name
FF 05	Lyric	Lyric text for karaoke
FF 06	Marker	Section marker (intro, chorus)
FF 20	MIDI Channel Prefix	Route following SysEx to channel
FF 2F 00	End of Track	Required at end of every track
FF 51 03	Set Tempo	3 bytes: microseconds per quarter note (e.g., 500000 = 120 BPM)
FF 54	SMPTE Offset	Start offset for SMPTE sync
FF 58	Time Signature	Numerator, denominator (power of 2), clocks per metronome, 32nds per quarter
FF 59	Key Signature	Key (0=C, 1=G..., -1=F...), mode (0=major, 1=minor)
FF 7F	Sequencer Specific	Application-specific data

System Exclusive (SysEx)

SysEx messages (0xF0 ... 0xF7) allow device-specific communication — sending patch data, tuning tables, custom parameters. SysEx IDs are registered with the MMA to avoid conflicts. Universal SysEx messages (0x7E and 0x7F IDs) provide standardized functions like Sample Dump, MIDI Show Control, and MIDI Machine Control.

MIDI Timing and Tempo

Tempo is stored in microseconds per quarter note (µs/quarter):

120 BPM = 500,000 µs/quarter
60 BPM = 1,000,000 µs/quarter
240 BPM = 250,000 µs/quarter

A Set Tempo event (FF 51 03 tt tt tt) can change tempo mid-file. With 480 PPQN and 500,000 µs/quarter, each tick = 500,000/480 ≈ 1,041.7 µs ≈ 1.04 ms.

MIDI timing has inherent imprecision due to the serial transmission delay (~320 µs per byte) and the sequencer's scheduling resolution. For video sync, SMPTE timecode (FF 54 meta event) is used.

MIDI File Tools and Conversion

# Play MIDI with TiMidity++ software synthesizer
timidity input.mid

# Play with FluidSynth and a SoundFont
fluidsynth -a alsa /usr/share/sounds/sf2/FluidR3_GM.sf2 input.mid

# Convert MIDI to audio with FluidSynth
fluidsynth -F output.wav /path/to/soundfont.sf2 input.mid
ffmpeg -i output.wav -c:a libmp3lame -q:a 2 output.mp3

# Convert MIDI to audio with TiMidity++
timidity input.mid -Ow -o output.wav

# Analyze MIDI file structure
python3 -c "import mido; mid = mido.MidiFile('input.mid'); print(mid)"

# Python: read all note events
python3 << 'EOF'
import mido
mid = mido.MidiFile('input.mid')
for track in mid.tracks:
    for msg in track:
        if msg.type in ['note_on', 'note_off']:
            print(msg)
EOF

# Merge multiple MIDI files
python3 -c "
import mido
mid1 = mido.MidiFile('track1.mid')
mid2 = mido.MidiFile('track2.mid')
merged = mido.MidiFile(type=1)
for track in mid1.tracks + mid2.tracks[1:]:
    merged.tracks.append(track)
merged.save('merged.mid')
"

MIDI in Modern Production

MIDI remains the universal language between:

DAWs: Ableton Live, Logic Pro, Pro Tools, FL Studio, Cubase, Reaper — all use MIDI internally
Hardware synthesizers: Roland, Korg, Moog, Sequential — all have MIDI IN/OUT
Software instruments: VST/AU/AAX plugins (Kontakt, Serum, Massive, Omnisphere) respond to MIDI
Controllers: Piano keyboards, pads (Maschine, Push), wind controllers, drum kits
Stage equipment: Lighting consoles (DMX via MIDI Show Control), mixers, effects units

MIDI 2.0 (released 2020) extends message resolution from 7-bit to 32-bit values (4.3 billion levels of expression vs. 128), adds per-note controllers, bidirectional property exchange, and profile configuration — while maintaining backward compatibility with MIDI 1.0.

Related conversions

Frequent conversions across the catalogue:

Frequently Asked Questions

MIDI files contain no audio samples — they store musical instructions (note numbers, timing, velocity) not sound. A 3-minute MIDI file is typically 10–100 KB. An equivalent MP3 is 3–10 MB. A MIDI file is like a sheet music score; an MP3 is like a recording of someone playing that score. The MIDI file must be played by a synthesizer to produce sound. The sound quality depends entirely on the synthesizer or SoundFont, not the MIDI file itself.

MIDI arquivos contain no audio samples — they store musical instructions (note numbers, timing, velocity) not sound. A 3-minute MIDI arquivo é tipicamente 10–100 KB. An equivalent MP3 is 3–10 MB. A MIDI arquivo é like a sheet music score; an MP3 is like a recording of someone reproduzindo that score. The MIDI arquivo must be played by a synthesizer to produce sound. The sound quality depends entirely no synthesizer ou SoundFont, not the MIDI arquivo itself.

MIDI Dateien contain no audio samples — they store musical instructions (note numbers, timing, velocity) not sound. A 3-minute MIDI Datei is typically 10–100 KB. An equivalent MP3 is 3–10 MB. A MIDI Datei is like a sheet music score; an MP3 is like a recording von someone playing that score. The MIDI Datei must be played by a synthesizer to produce sound. The sound quality depends entirely auf dem synthesizer oder SoundFont, not the MIDI Datei itself.

MIDI archivos contain no audio samples — they store musical instructions (note numbers, timing, velocity) not sound. A 3-minute MIDI archivo is typically 10–100 KB. An equivalent MP3 is 3–10 MB. A MIDI archivo is like a sheet music score; an MP3 is like a recording de someone playing that score. The MIDI archivo must be played by a synthesizer to produce sound. The sound quality depends entirely en el synthesizer o SoundFont, not the MIDI archivo itself.

On KaijuConverter every file is processed inside an isolated container, encrypted in transit (TLS 1.3) and at rest, and automatically deleted after 60 minutes with multi-pass overwrite. We never train on, share, or analyze user content. For maximum privacy on extremely sensitive material, prefer offline tools (ImageMagick, FFmpeg, LibreOffice) that you control end-to-end.

MIDI must be rendered through a synthesizer to produce audio. Use FluidSynth with a SoundFont: `fluidsynth -F output.wav soundfont.sf2 input.mid` then convert WAV to MP3 with FFmpeg. TiMidity++ is another option: `timidity input.mid -Ow -o output.wav`. For higher quality, use a DAW (Ableton, Logic, GarageBand) with premium virtual instruments — load the MIDI, assign high-quality VST instruments, and export/bounce to audio. The SoundFont or VST instrument quality determines the final audio quality, not the MIDI file.

MIDI must be rendered através de a synthesizer to produce audio. usar FluidSynth com a SoundFont: `fluidsynth -F output.wav soundfont.sf2 input.mid` then converter WAV to MP3 com FFmpeg. TiMidity++ is another option: `timidity input.mid -Ow -o output.wav`. para altaer quality, usar a DAW (Ableton, Logic, GarageBand) com premium virtual instruments — load the MIDI, assign alta-quality VST instruments, e export/bounce to audio. The SoundFont ou VST instrument quality determines the final qualidade do áudio, not the MIDI file.

MIDI must be rendered durch a synthesizer to produce audio. verwenden FluidSynth mit a SoundFont: `fluidsynth -F output.wav soundfont.sf2 input.mid` then umwandeln WAV to MP3 mit FFmpeg. TiMidity++ is another option: `timidity input.mid -Ow -o output.wav`. für hocher quality, verwenden a DAW (Ableton, Logic, GarageBand) mit premium virtual instruments — load the MIDI, assign hoch-quality VST instruments, und export/bounce to audio. The SoundFont oder VST instrument quality determines the final Audioqualität, not the MIDI file.

MIDI must be rendered a través de a synthesizer to produce audio. usar FluidSynth con a SoundFont: `fluidsynth -F output.wav soundfont.sf2 input.mid` then convertir WAV to MP3 con FFmpeg. TiMidity++ is another option: `timidity input.mid -Ow -o output.wav`. para altaer quality, usar a DAW (Ableton, Logic, GarageBand) con premium virtual instruments — load the MIDI, assign alta-quality VST instruments, y export/bounce to audio. The SoundFont o VST instrument quality determines the final calidad de audio, not the MIDI file.

For 95% of use cases, yes — server-side ImageMagick, FFmpeg and LibreOffice produce identical output to the same tools on your laptop. Desktop software wins for: extremely large files (multi-GB), batch jobs of thousands of files, scripted pipelines, or content too sensitive to upload. KaijuConverter caps at 25 MB per file on the free tier (up to 2 GB on paid plans).

General MIDI (GM) standardizes which instrument plays on which program change number — program 0 is always Grand Piano, program 40 is always Violin, channel 10 is always drums. However, GM only standardizes which instrument is used, not how it sounds. A GM piano sample on a cheap soundcard sounds completely different from a high-quality Steinway sample. This is why the same MIDI file sounds different on every device. Professional MIDI productions embed SysEx data to configure specific synthesizers, or use a specific GM SoundFont (.SF2) for consistent playback.

General MIDI (GM) padrãoizes which instrument plays on which program change number — program 0 is always Grand PiaNão, program 40 is always Violin, channel 10 is always drums. Porém, GM only padrãoizes which instrument is used, not how it sounds. A GM piano sample em um cheap soundcard sounds completely diferente de a alta-quality Steinway sample. This is why the same MIDI arquivo sounds different on every device. profissional MIDI productions embed SysEx data to configure specific synthesizers, ou usar a specific GM SoundFont (.SF2) para consistent playback.

General MIDI (GM) Standardizes which instrument plays on which program change number — program 0 is always Grand PiaNein, program 40 is always Violin, channel 10 is always drums. Jedoch, GM only Standardizes which instrument is used, not how it sounds. A GM piano sample auf einem cheap soundcard sounds completely anders als a hoch-quality Steinway sample. This is why the same MIDI Datei sounds different on every device. professionell MIDI productions embed SysEx data to configure specific synthesizers, oder verwenden a specific GM SoundFont (.SF2) für consistent playback.

General MIDI (GM) estándarizes which instrument plays on which program change number — program 0 is always Grand PiaNo, program 40 is always Violin, channel 10 is always drums. Sin embargo, GM only estándarizes which instrument is used, not how it sounds. A GM piano sample en un cheap soundcard sounds completely diferente a a alta-quality Steinway sample. This is why the same MIDI archivo sounds different on every device. profesional MIDI productions embed SysEx data to configure specific synthesizers, o usar a specific GM SoundFont (.SF2) para consistent playback.

Most format conversions are lossy by design — JPG, MP3, MP4, WebP all discard perceptual data to save bytes. Going through a lossy intermediate compounds the loss. To minimize visible/audible drift: convert from the original master, choose a higher quality setting, and avoid converting back and forth between lossy formats.

MIDI 2.0 (standardized 2020, hardware arriving 2022–2024) dramatically upgrades MIDI 1.0's 7-bit (128 values) resolution to 32-bit (4.3 billion values) for note velocity, pitch bend, and controllers — enabling micro-pitch tuning, nuanced expression, and per-note dynamics impossible in MIDI 1.0. It adds bidirectional communication (instruments can tell the host what capabilities they have), channel-based profile configuration, and property exchange. MIDI 2.0 is backward compatible — MIDI 1.0 devices and files still work. For most users, MIDI 2.0 is still emerging. MIDI 1.0 files (.mid) remain the universal standard for distribution.

MIDI 2.0 (standardized 2020, hardware arriving 2022–2024) dramatically upgrades MIDI 1.0's 7-bit (128 values) resolução to 32-bit (4.3 billion values) para note velocity, pitch bend, e controllers — enabling micro-pitch tuning, nuanced expression, e per-note dynamics impossible in MIDI 1.0. It adds bidirectional communication (instruments can tell the host what capabilities they have), channel-based profile configuration, e property exchange. MIDI 2.0 is backward compatível — MIDI 1.0 devices e arquivos still work. para most users, MIDI 2.0 is still emerging. MIDI 1.0 arquivos (.mid) remain the universal padrão para distribution.

MIDI 2.0 (standardized 2020, hardware arriving 2022–2024) dramatically upgrades MIDI 1.0's 7-bit (128 values) Auflösung to 32-bit (4.3 billion values) für note velocity, pitch bend, und controllers — enabling micro-pitch tuning, nuanced expression, und per-note dynamics impossible in MIDI 1.0. It adds bidirectional communication (instruments can tell the host what capabilities they have), channel-based profile configuration, und property exchange. MIDI 2.0 is backward kompatibel — MIDI 1.0 devices und Dateien still work. für most users, MIDI 2.0 is still emerging. MIDI 1.0 Dateien (.mid) remain the universell Standard für distribution.

MIDI 2.0 (standardized 2020, hardware arriving 2022–2024) dramatically upgrades MIDI 1.0's 7-bit (128 values) resolución to 32-bit (4.3 billion values) para note velocity, pitch bend, y controllers — enabling micro-pitch tuning, nuanced expression, y per-note dynamics impossible in MIDI 1.0. It adds bidirectional communication (instruments can tell the host what capabilities they have), channel-based profile configuration, y property exchange. MIDI 2.0 is backward compatible — MIDI 1.0 devices y archivos still work. para most users, MIDI 2.0 is still emerging. MIDI 1.0 archivos (.mid) remain the universal estándar para distribution.

Yes — KaijuConverter accepts multiple files in a single drop and returns a ZIP. For very large batches (thousands of files) consider command-line tools or our API: <code>find . -name "*.heic" -exec magick {} {.}.jpg \;</code> or similar one-liners scale to millions of files when run locally.