Skip to content

Latest commit

 

History

History
511 lines (365 loc) · 13.1 KB

RESEARCH_SUMMARY.md

File metadata and controls

511 lines (365 loc) · 13.1 KB
title created modified description tags
Logic Pro ProjectData Binary Format — Research Summary
2026-02-04
2026-02-28
Successfully reverse-engineered significant portions of Logic Pro's proprietary ProjectData binary format through systematic analysis. Extracted plugin configurations, preset data, and identified the...
music

Logic Pro ProjectData Binary Format — Research Summary

Logic Pro ProjectData Binary Format - Research Summary

Executive Summary

Successfully reverse-engineered significant portions of Logic Pro's proprietary ProjectData binary format through systematic analysis. Extracted plugin configurations, preset data, and identified the chunk-based file structure. Track names remain partially elusive but significant progress has been made.

Key Discoveries

1. File Structure

File Header:

  • Magic bytes: 23 47 c0 ab (appears to be file signature)
  • Contains early reference to "gnoS" (Song backwards?)
  • Version information embedded as text strings

Overall Organization:

  • Chunk-based binary format (similar to IFF/RIFF)
  • 801 total chunks in analyzed file (2.8 MB)
  • Hierarchical structure with nested data

2. Chunk Types Identified (4-byte reversed ASCII markers)

Bytes Reversed Count Purpose Status
karT Trak 320 Track definition ✅ Identified
qeSM MSeq 169 MIDI sequence data ✅ Identified
qSvE EvSq 169 Event sequence (automation) ✅ Identified
gRuA AuRg 38 Audio region ✅ Identified
tSxT TxSt 32 Text/Score notation styles ✅ Decoded
lFuA AuFl 23 Audio file reference ✅ Identified
LFUA AUFL 23 Audio file (variant) ✅ Identified
PMOC COMP 23 Comping/take data ✅ Identified
MroC CorM 2 Core MIDI? ⚠️ Partial
tSnI InSt 1 Instrument definition ✅ Identified
snrT Trns 1 Transform/transpose ⚠️ Partial

Note: The reverse-naming pattern (FourCC codes read backwards) is consistent throughout.

3. Data Encoding Methods

String Encoding (Multiple formats coexist):

  1. Length-prefixed (Pascal):

    [1-byte length][string data]

  2. 2-byte length (Big-Endian):

    [2-byte BE length][string data]

  3. 4-byte length (Big-Endian):

    [4-byte BE length][string data]

    Used for large JSON blocks

  4. Null-terminated:

    [string data]\x00

Numeric Data:

  • Float32: BPM values, plugin parameters (both BE and LE found)
  • Int32: Sample positions, region lengths, counts (mostly BE)
  • Endianness: Mixed, but Big-Endian appears dominant

4. Embedded JSON Data ✅ MAJOR FIND

Discovery: Large sections contain complete JSON objects for Session Players instruments

Structure:

{
  "Preset": {
    "Name": "Sweet Memories",
    "CharacterIdentifier": "Acoustic Piano - Strummed",
    "PresetIdentifier": "Acoustic Piano - Strummed - Sweet Memories.dpst",
    "Type": "TypeFactoryPreset",
    "Parameters": {
      "intensity": 71,
      "dynamics": 119,
      "humanize": 37,
      "variation": 1,
      "fillsAmount": 69,
      "rComp": 42,
      "mComp": 3,
      ...
    }
  },
  "GeneratorMemento": {
    "Seeds": {"0": 1763219906, ...},
    "LastGenerateLength": 32,
    "MementoParameters": {
      "regionStartParams": "...",
      "regionEndParams": "...",
      "humanizeEndStatus": "...",
      ...
    }
  },
  "RegionType": "Type_AcousticPianoV2",
  "FollowTrackUniqueID": "43870E44-C22E-11F0-968B-D17F5F851CEB"
}

Decoded Parameters:

  • Intensity: Musical intensity (0-127)
  • Dynamics: Velocity range (0-127)
  • Humanize: Timing variation (0-100)
  • Variation: Pattern variation (1-4)
  • fillsAmount: Drum fill density
  • rComp: Rhythm complexity
  • mComp: Melodic complexity
  • pushPull: Timing feel (-50 to +50)
  • swing: Swing amount (0-100)

5. Plugin/Instrument Data Extracted

Synthesizers:

  • Alchemy:

    • Oscillator shapes: Sine, Saw, Triangle, Square
    • LFO shapes: Ramp Up, Ramp Down, Random, Sine
    • Formant filters: Vowel shapes (A, E, I, U)
    • Noise sources: White, Radio
    • Window functions: Tukey
    • Library paths successfully extracted

  • Sampler/Quick Sampler:

    • References found but minimal data

Session Players (with full preset data):

  1. Acoustic Piano - Strummed:

    • Preset: "Sweet Memories"
    • Type: Type_AcousticPianoV2
    • Voicing styles, positions, humanization all decoded

  2. Electric Bass - Indie Disco:

    • Preset: "Night Flight"
    • Type: Type_ElectricBassV2
    • Bass type, slides, dead notes, playing position decoded

  3. Electronic Drummer - Synthpop:

    • Preset: "Hit Factory"
    • Type: Type_ElectronicDrummerV2
    • Kit pieces (a1, a2, a3, b1, b2, b3, c1, c2, c3) complexity mapped
    • Accent variation, phrase variation extracted

6. Text/Score Notation Styles (tSxT chunks)

Successfully decoded 32 text style definitions:

  • Plain Text
  • Page Number
  • Bar Number
  • Instrument Name
  • Tuplet
  • Repeat Ending
  • Chord Root/Extensions
  • Multiple Rest
  • Tablature
  • Tempo Symbol
  • Octave Symbol
  • Note Head
  • Guitar Grid Fingering
  • Guitar Marking
  • Fingering
  • Reserved styles (1-3, numbered slots)

Each with font specifications (Times-Italic found).

7. Audio File References

Format: Relative and absolute paths

  • Alchemy library: Alchemy/Libraries/WaveOsc/Basic/Sine.raw
  • System paths: /Library/Application Support/Logic/Alchemy/...
  • Apple Loops: Library/Audio/Apple Loops/Apple/21 Prismatica...

8. Automation/Event Data

MIDI Sequences (qeSM):

  • Contains note data
  • "Untitle" strings found (possibly default region names)
  • 169 sequences found

Event Sequences (qSvE):

  • Automation curves
  • Parameter modulation
  • 169 instances (matches MIDI sequences 1:1)

9. Track Names - ONGOING CHALLENGE

Current Status: Partially solved

What We Know:

  • Track markers (karT) do NOT have immediately adjacent names
  • Generic strings found: "Audio 1", "Audio 2", etc.
  • Session Player region names stored in JSON
  • Score notation names found in tSxT chunks

Hypotheses:

  1. Names stored in separate index table (not yet located)
  2. Names use indirect references (offsets/pointers)
  3. May be in file header or footer section
  4. Could be in parent container we haven't identified

Evidence of Names:

  • "Score Settings" found near tSnI instrument chunk
  • "Plain Text", "Page Number" etc. are style names, not track names
  • "Untitle" appears near qeSM markers (might be default MIDI region name)

Technical Analysis

Chunk Structure Pattern

Standard chunk appears to follow:

Offset  Bytes       Description
+0      [4 bytes]   Chunk type (reversed FourCC)
+4      [4 bytes]   Flags or sub-type (BE int32)
+8      [4 bytes]   Unknown (often 0x00000000)
+12     [4 bytes]   Unknown (often 0x0000FFFF or 0xFFFFFFFF)
+16     [variable]  Payload data

Observed Header Integers (Big-Endian):

  • karT: [83891968, 0, 65535, 4294967295]
  • qeSM: [83886336, 0, 65535, 4294967295]
  • qSvE: [16777472, 0, 13824, 65535]
  • tSnI: [16777728, 0, 65535, 4294967295]

File Size & Chunk Distribution

Example Project (Hollywood Solaris):

  • Total size: 2,801,481 bytes (2.8 MB)
  • 801 chunks
  • Average chunk size: ~3,497 bytes
  • Largest chunks: JSON preset data (several KB)
  • Smallest chunks: 36 bytes (minimal karT markers)

Performance Parameters Decoded

Drummer Parameters:

  • Kit piece complexity: a1-a3 (hi-hats, rides), b1-b3 (snares, claps), c1-c3 (kicks, toms)
  • Each with min/max complexity values
  • System variation (1-4): Overall pattern style
  • Phrase variation (0-100): Pattern diversity
  • Accent variation (0-4): Dynamic accents

Bass Parameters:

  • Position (1-100): Fretboard position
  • Slides amount (0-100)
  • Dead notes (0-100)
  • Buzz trills (0-100)
  • Blue-syness (0-100): Blues inflection
  • Riffiness (1-5): Pattern complexity
  • Playing position: Pickup vs bridge

Piano Parameters:

  • Upper/lower voicing style (1-5)
  • Upper/lower playing style (1-5)
  • Grace notes amount (0-100)
  • Arp direction (0-3)
  • Emphasis (-50 to +50)

Tools Developed

  1. logic_project_analyzer_enhanced.py:

    • Main analyzer with track name extraction
    • Generates markdown reports
    • Handles MetaData.plist + ProjectData

  2. binary_format_analyzer.py:

    • Magic marker detection
    • String extraction (multiple methods)
    • Repeating structure identification
    • Hex context analysis

  3. extract_plugin_data.py:

    • JSON object extraction
    • Plugin name detection
    • Preset configuration decoding
    • Audio file reference extraction

  4. hex_dump_analyzer.py:

    • Annotated hex dumps
    • Context analysis around markers
    • Multiple string extraction attempts
    • Numeric data patterns

  5. chunk_structure_analyzer.py:

    • Full file structure mapping
    • Chunk hierarchy visualization
    • Header parsing
    • Track name candidate analysis

Remaining Mysteries

High Priority:

  1. Track Names: Primary location still unknown
  2. Chunk Size Field: Format unclear (not simple 4-byte size)
  3. Chunk Hierarchy: Parent-child relationships not fully mapped
  4. Compression: PMOC (COMP) chunks may use compression (not yet tested)

Medium Priority:

  1. Plugin States: AU/VST plugin data encoding
  2. Mixer Settings: Channel strip, routing, sends
  3. Smart Controls: Mapping and configuration
  4. Flex Time/Pitch: Audio editing metadata

Low Priority:

  1. Undo History: Project file backups format
  2. Color Coding: Track/region colors
  3. Folder Tracks: Grouping structure

Next Steps

Immediate Actions:

  1. Search for Name Table:

    • Analyze file header more thoroughly
    • Look for offset pointer tables
    • Check file footer sections

  2. Test Compression:

    • Try zlib decompression on PMOC chunks
    • Check for other compression algorithms

  3. Compare Projects:

    • Analyze multiple projects with known track names
    • Diff files to isolate name storage location

Medium-term:

  1. Build Parser Library:

    • Python module for ProjectData parsing
    • Clean API for extracting data
    • Integration with existing analyzer

  2. Document AU/VST Formats:

    • Reverse engineer plugin state chunks
    • Map common plugins (EQ, compressor, etc.)

Long-term:

  1. Complete Format Specification:
    • Document all chunk types
    • Create format reference guide
    • Enable third-party tools

Conclusions

Successfully Decoded:

✅ File structure (chunk-based) ✅ Chunk markers and types ✅ Session Players presets (full parameter sets) ✅ Alchemy synthesizer configurations ✅ Score notation styles ✅ Audio file references ✅ MIDI/automation sequences (structure) ✅ JSON-embedded configurations

Partially Decoded:

⚠️ Track names (generic names found, custom names elusive) ⚠️ Chunk hierarchy ⚠️ Some marker types (MroC, snrT)

Not Yet Decoded:

❌ AU/VST plugin states ❌ Mixer configurations ❌ Smart Controls ❌ Flex editing data ❌ Compression schemes

Research Value

This reverse engineering effort provides:

  1. Enhanced Project Analysis:

    • Extract detailed instrument configurations
    • Analyze musical parameters across projects
    • Track plugin usage patterns

  2. Workflow Insights:

    • Understand Session Players generation
    • Decode preset parameters for learning
    • Extract musical metadata

  3. Future Tool Development:

    • Foundation for ProjectData parsers
    • Potential for project conversion tools
    • Cross-DAW project migration

  4. Format Documentation:

    • First comprehensive analysis of ProjectData format
    • Reference for future research
    • Enables community tools

Research Date: December 5, 2025 Logic Pro Version: 11.x (detected from files) Files Analyzed: 3 projects, ~2-3 MB each Format Coverage: ~60% understood, 40% remaining

Tools Available At: This repository - see README.md for usage instructions

Sample Output

From analyzing "Example Project.logicx":

Plugins Detected:

  • Alchemy (synthesizer with 28 library references)
  • Sampler
  • Q-Sampler (Quick Sampler)

Presets Found:

  • Acoustic Piano - "Sweet Memories" (Type_AcousticPianoV2)
    • Intensity: 71, Dynamics: 119, Humanize: 37
  • Electric Bass - "Night Flight" (Type_ElectricBassV2)
    • Intensity: 79, Dynamics: 100, Riffiness: 3
  • Electronic Drummer - "Hit Factory" (Type_ElectronicDrummerV2)
    • Intensity: 97.3, Fill Amount: 30.5

File Statistics:

  • Size: 2,801,481 bytes
  • Total Chunks: 801
  • Tracks: 320
  • MIDI Sequences: 169
  • Audio Regions: 38
  • Text Styles: 32

References

  • No official documentation exists for this format
  • Format appears proprietary to Apple Logic Pro
  • Similar chunk-based formats: IFF, RIFF, AIFF
  • Endianness: Primarily Big-Endian (network byte order)