AudioSync Pro

Multi-device audio/video synchronization with FFT cross-correlation, automatic clock drift detection, and NLE timeline export. Built with Rust and Tauri v2 for native performance.

Free and open source. Website | Download | Report Issue

What It Does

You record a session with multiple devices — Camera A rolls continuously, Camera B takes short B-roll clips throughout, and a Zoom recorder captures audio in several segments. All devices were running at the same time but started and stopped independently.

AudioSync Pro finds exactly where each clip sits on a shared timeline using audio cross-correlation, then exports one perfectly synced audio file per device. Drop the exports into any DAW or video editor and they line up.

Key Features

FFT cross-correlation — Sample-accurate sync across devices
Clock drift detection — Measures and corrects drift (ppm) between devices
Auto-grouping — Files grouped by device name automatically
NLE timeline export — FCPXML (Final Cut Pro / DaVinci Resolve) and EDL (Premiere / Avid)
Multiple formats — WAV, AIFF, FLAC, MP3 (16/24/32-bit)
Video support — Extract audio from MP4, MOV, MKV, AVI, etc. via ffmpeg
Cloud save/load — Save projects to the cloud via Keyhan Studio account (optional)
Timeline sharing — Share synced timelines via a public link with interactive viewer
Cross-platform — macOS, Windows, Linux

Three Ways to Use AudioSync Pro

1. Desktop App (Tauri v2 + Vue 3)

The full GUI experience with waveform visualization, drag-and-drop, and real-time progress.

# Build and run
npm install
cargo tauri dev

Features: Glassmorphism UI, Canvas waveform timeline, resizable panels, file drag-and-drop, native menus, keyboard shortcuts (Cmd+O/S/R/E/D), progress dialogs, drift measurement tool, project save/load.

2. Rust CLI

Headless command-line tool for servers, pipelines, and automation.

# Build the CLI
cargo build --release -p audiosync-cli

# Analyze files
./target/release/audiosync analyze CamA_001.mp4 CamA_002.mp4 Zoom_001.wav --json

# Sync and export
./target/release/audiosync sync *.mp4 *.wav -o ./output --format wav --bit-depth 24

# Measure clock drift between two files
./target/release/audiosync drift -r reference.wav -t target.wav

# Show file info and auto-grouping
./target/release/audiosync info *.mp4 *.wav

Flags: --json for pipe-friendly output, --max-offset to constrain search, --no-drift-correction, --save for project files, --fcpxml / --edl for timeline export.

3. Python CLI (Legacy)

The original Python implementation is preserved in the python/ directory for users who prefer it.

cd python/
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

# Analyze and sync
python cli.py analyze CamA_001.mp4 CamA_002.mp4 Zoom_001.wav
python cli.py sync CamA_001.mp4 CamA_002.mp4 Zoom_001.wav -o ./output

See python/README.md for full documentation.

Architecture

AudioSyncPro/
├── audiosync-core/       # Rust library — engine, models, audio I/O
│   └── src/
│       ├── models.rs         # Clip, Track, SyncConfig, SyncResult
│       ├── audio_io.rs       # Symphonia + ffmpeg loading, rubato resampling, hound export
│       ├── engine.rs         # FFT cross-correlation, drift detection, analysis pipeline
│       ├── grouping.rs       # Auto-group files by device name
│       ├── metadata.rs       # ffprobe creation timestamps
│       ├── project_io.rs     # JSON project save/load
│       ├── timeline_export.rs# FCPXML v1.11 + EDL (CMX 3600)
│       └── cloud.rs          # Cloud API client (future)
├── audiosync-cli/        # Rust CLI binary
├── src-tauri/            # Tauri v2 desktop app (Rust backend)
│   └── src/
│       ├── commands.rs       # IPC bridge (15 commands)
│       ├── menu.rs           # Native app menu
│       └── lib.rs            # App entry, plugins, state
├── src/                  # Vue 3 frontend
│   ├── composables/
│   │   ├── useAudioSync.js   # Central state + Tauri invoke wrappers
│   │   ├── useAuth.js        # Device code OAuth + JWT session management
│   │   ├── useCloud.js       # Cloud project CRUD + timeline sharing
│   │   └── useToast.js       # Toast notification system
│   ├── components/
│   │   ├── MainLayout.vue    # App shell, toolbar, shortcuts, drag-drop
│   │   ├── WorkflowBar.vue   # 3-step workflow indicator
│   │   ├── TrackPanel.vue    # Track list sidebar
│   │   ├── TrackCard.vue     # Per-track card with clip list
│   │   ├── WaveformCanvas.vue# Canvas 2D timeline with waveform peaks
│   │   ├── ResizeSplitter.vue# Draggable panel divider
│   │   ├── LoginDialog.vue   # Device code login flow
│   │   ├── CloudProjectsDialog.vue  # Cloud project list + save/load
│   │   ├── ShareDialog.vue   # Timeline sharing via link
│   │   ├── ProcessingDialog.vue
│   │   ├── ExportDialog.vue
│   │   ├── DriftFixDialog.vue
│   │   ├── AboutDialog.vue
│   │   └── ToastNotification.vue
│   └── styles/
│       ├── main.css          # Glassmorphism design system
│       └── animations.css    # Transitions and keyframes
├── python/               # Legacy Python implementation (preserved)
├── website/              # Marketing website
├── .github/workflows/    # CI + Release pipelines
├── Cargo.toml            # Rust workspace
└── package.json          # Node.js (Vue + Vite)

Algorithm

The analysis engine operates in 8 phases at 8 kHz mono:

Sort clips by creation timestamp (ffprobe metadata)
Select reference track (widest time coverage or longest duration)
Build reference timeline from metadata gaps between clips
Pass 1: FFT cross-correlation of each non-reference clip against the reference
Pass 2: Enhanced timeline retry for low-confidence clips (stitches all placed clips)
Metadata fallback for clips that still can't be matched
Normalize timeline so the earliest offset is zero
Drift detection via windowed cross-correlation + linear regression

Confidence metric: peak / mean ratio of the correlation (>3.0 = good match).

Drift measurement: Windowed cross-correlation at 30s intervals with 15s stride, sub-sample parabolic interpolation, linear regression of offsets → drift in ppm.

Development

Prerequisites

Rust (stable) — curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
Node.js 20+ — brew install node or nodejs.org
ffmpeg — brew install ffmpeg (required for video files)

Build

# Install frontend dependencies
npm install

# Run tests
cargo test --workspace

# Development mode (hot-reload)
cargo tauri dev

# Production build
cargo tauri build

# CLI only (no GUI dependencies)
cargo build --release -p audiosync-cli

Test

# All tests (51 total: 41 unit + 9 CLI integration + 1 doctest)
cargo test --workspace

# Core library only
cargo test -p audiosync-core

# CLI integration tests
cargo test -p audiosync-cli

Feature Comparison

| Feature | Desktop App | Rust CLI | Python CLI | |---------|:-----------:|:--------:|:----------:| | FFT cross-correlation | Yes | Yes | Yes | | Clock drift detection | Yes | Yes | Yes | | Drift correction | Yes | Yes | Yes | | Auto-grouping | Yes | Yes | Yes | | WAV/AIFF/FLAC/MP3 export | Yes | Yes | Yes | | FCPXML timeline export | Yes | Yes | No | | EDL timeline export | Yes | Yes | No | | Waveform visualization | Yes | -- | -- | | Drag-and-drop | Yes | -- | -- | | Project save/load | Yes | Yes | No | | Cloud save/load | Yes | -- | -- | | Timeline sharing | Yes | -- | -- | | JSON output | -- | Yes | Yes | | Headless/server use | -- | Yes | Yes | | GUI | Tauri + Vue 3 | -- | PyQt6 (legacy) |

Migration from v2.x (Python)

AudioSync Pro v3.0 is a ground-up rewrite in Rust. The algorithm is identical but the implementation is new.

What changed in v3.0:

Python/PyQt6 → Rust/Tauri v2 + Vue 3
soundfile + scipy → symphonia + rustfft + rubato
opentimelineio → native FCPXML/EDL generation
Single binary → Cargo workspace with 3 crates

What's new in v3.1.0:

Cloud save/load via Keyhan Studio account (optional, app works fully offline)
Timeline sharing via public link with interactive waveform viewer
Fixed drag-and-drop for Tauri v2 native file handling
Fixed intra-track clip overlap enforcement
Fixed FCPXML gap elements for DaVinci Resolve compatibility

What stays:

The Python implementation in python/ with its own CLI
The AudioSync.js Max for Live device
The website
GPL-3.0 license

Breaking changes:

Project file format is v2 (JSON) — not compatible with v1
CLI syntax changed: python cli.py analyze → audiosync analyze

License

GPL-3.0 — see LICENSE.

Created by Keyhan.

Audiosync

Install / Use

README