EAR Audio Preview

A professional-grade audio analysis tool in your IDE.
The studio-grade audio analysis, comparison, and monitoring suite—built right into VS Code & Cursor.

English · 简体中文

Quick Start · Workflows · What's New (v0.6) · Why EAR? · Architecture · Development

[!TIP] Engineering Deep Dives: Curious about how we achieve real-time DSP inside an editor? Explore our Technical Blog, covering everything from PAZ-inspired CQT spectrum design to anisotropic WebGL2 spectrogram rendering.

⚡ Overview

Stop switching between your code editor and heavy external DAWs.

EAR Audio Preview transforms VS Code and Cursor into a high-precision acoustic workstation. Whether you are evaluating generative AI audio models, curating speech datasets, debugging DSP pipelines, or reviewing mix revisions, EAR delivers real-time waveforms, GPU-accelerated spectrograms, live stereo imaging, and EBU R128 loudness metering directly where you build.

🚀 Quick Start

1. Installation

Search for EAR Audio Preview in your editor's Extensions panel, or install instantly via CLI:

code --install-extension Eps-Acoustic-Revolution-Lab.ear-audio-preview

2. Master EAR in 60 Seconds

1. Open & Audition: Click any audio file and press Space to play. (If another extension takes over, right-click the tab and select Reopen Editor With… → Audio Preview).
2. Inspect the Details: Drag across the STFT plot to zoom into micro-transients. Switch to Live Spec for real-time frequency & stereo-field telemetry, or open Loudness to compute LUFS, pitch ($F_0$), and onset timelines.
3. Multi-Track Comparison: Select 2 to 8 mono/stereo files in the Explorer, right-click, and choose EAR Audio: Compare Files. Play them in lockstep and toggle per-track mute/focus to compare what you hear.

[!NOTE] Pro Controls & Formats: An interactive onboarding guide greets you on first launch (replay anytime via ? in the toolbar). In comparison view, use Ctrl/Cmd+Drag to zoom time, Shift+Drag to zoom frequency, and Double-Click to reset. Right-click any plot for layout and fullscreen toggles. Supported Formats: WAV · MP3 · AAC · OGG · FLAC · OPUS · M4A · SPH 📖 Full User Guide · ⚙️ Settings Reference

🎬 See It in Action

Compare Model Outputs, References & Mix Revisions

Load up to eight recordings on a unified timeline. Drag once to inspect the exact same time-frequency region across every spectrogram simultaneously. Switch to Live Spec for color-coded spectrum overlays, goniometer stereo fields, and phase-correlation telemetry. Need a visual baseline? Keep your reference track visible in analysis while muting it during playback.

Check Dynamics & Calibrate Your Listening

Seamlessly transition from casual playback to rigorous broadcast-grade diagnostics—measuring integrated LUFS, True Peak, fundamental frequency ($F_0$), and onset curves. Open the floating Monitor panel to isolate L / R / M / S channels or solo specific frequency bands, then apply an AutoEq headphone correction profile and toggle bypass to hear your audio on a flat reference response.

A Complete Acoustic Inspection Suite

What's new in v0.6.0

• Align before comparing. Estimate a constant integer-sample delay between two same-rate files, review it, then apply or reset the session copy across playback, analysis and export. Polarity compensation is optional; source files stay unchanged.

• Eight files, one workspace. Parallel playback, linked waveform/STFT zoom and independent track focus, mute and analysis controls.

• Live comparison at a glance. Overlay FFT/CQT spectra, stereo fields and phase correlation; compare LUFS, F0 and onset timelines.

• Match levels and inspect differences. Match shared-region RMS without boosting peaks; measure same-rate residual RMS and peak against a reference.

• Export the mix you choose. Save a selected region with configurable sample rate, channels and PCM or float32 WAV encoding, with optional headphone EQ.

• A smoother daily workflow. Unified controls, a floating Monitor with headphone correction, an interactive comparison guide and a 48% smaller package.

• More reliable diagnostics. Correct sound-field direction and polarity handling, bound polar interpolation, and recover safely from malformed saved headphone EQ.

Read the technical blog: Why Audio Alignment Fails (中文).

📜 Explore Full Release History (CHANGELOG)

🏆 Why EAR?

Most editor plugins stop at basic waveform playback. EAR Audio Preview is built for engineers and researchers who need to hear and see every artifact across multiple renders simultaneously.

| Core Capability | EAR Audio Preview | audio-preview | Audio Player | audioscope | | :--- | :---: | :---: | :---: | :---: | | Multi-Track Parallel Comparison | ✅ 8 Tracks · Linked Zoom | Single file | Sequential only | 2-file A/B · Linked | | Time-Frequency Spectrogram (STFT) | ✅ WebGL2 · Anisotropic | ✅ Basic | Waveform only | ✅ Basic | | Real-Time Spectrum Analyzer | ✅ FFT + CQT · Multi-Overlay | — | — | — | | Stereo Imaging & Phase Diagnostics
_{3-mode Goniometer, Freq Correlation & M/S Meters} | ✅ Full Suite | — | — | — | | Loudness & Acoustic Feature Timelines
_{EBU R128 (LUFS / LRA / True Peak), Pitch ($F_0$) & Onsets} | ✅ Full Suite | — | — | LUFS / Peak only | | Sample-Accurate Alignment & Level Match
_{Auto delay/polarity compensation, RMS match & residual diff} | ✅ Full Suite | — | — | ✅ | | Studio Monitoring Matrix
_{L/R/M/S channel routing & 5-band crossover solo} | ✅ Full Matrix | — | — | — | | Reference Headphone Calibration
_{Built-in AutoEq profile search, import & instant bypass} | ✅ Built-in AutoEq | — | — | — | | Precision Region Bounce & Mix Export
_{Custom sample rate, bit depth, mix bounce & optional EQ bake} | ✅ Stem / Mix / EQ Bake | Single clip | — | Single clip | | Decoupled Analysis & Listening Buses
_{Solo bands or apply EQ without corrupting objective meters} | ✅ Isolated Buses | — | — | — | ✅ Supported · — Not available in compared version. See detailed version & feature notes.

Decoupled Analysis & Monitoring: Keep a reference track muted in your headphones while still watching its live spectrum on screen. Channel routing, band soloing, and headphone EQ adjust what you hear without corrupting your underlying measurements. When Align is applied, all pipelines seamlessly lock onto the time-aligned session copy.

⚡ Blazing Fast to Inspect. Featherlight to Install.

📊 Full Benchmark against 6 Audio Extensions · 🔬 Benchmark Methodology & Suite

🔧 Under the Hood

🎨 Anisotropic Spectrograms that Keep Transients Sharp

Standard bilinear filtering blurs sharp percussive attacks. EAR's custom WebGL2 renderer decouples the axes: preserving crisp, unblurred temporal transients along the X-axis while applying smooth interpolation along the frequency Y-axis to keep harmonic contours clean. Combined with adaptive FFT sizing and multiple window functions, you get uncompromising clarity for both micro-events and sustained tones.

👉 Read: How the Anisotropic Renderer Works

🎹 A Spectrum Analyzer Tuned for Human Hearing

Switch effortlessly between standard linear/log FFT and a PAZ-inspired Constant-Q Transform (CQT) analyzer featuring variable-Q filter banks, deep low-frequency resolution, adjustable spectral tilt, and configurable ballistic release—ensuring the visual display stays musical and readable in real time.

👉 Read: CQT Spectrum Analyzer Design Notes

⏱️ Reversible Audio Alignment

Click Align to automatically estimate integer-sample delay and polarity flips between two same-rate recordings (searching a ±2s window across the first 60s). Once applied, playback, visual plots, and exports all reference the shifted session buffer—while a single click on Reset restores the bit-exact original. (Note: Fractional subsample delays and clock drift are intentionally out of scope).

👉 Read: Alignment DSP, Edge Cases & Ablation · Usage Guide

🧭 Shared Timeline, Source-Aware DSP Architecture

In multi-file comparison, each track is analyzed natively at its decoded sample rate and projected onto unified time and frequency coordinates with a shared absolute dBFS color scale. Because Source Analysis, Live Monitoring, and Export run on isolated DSP buses, tweaking your monitoring EQ or soloing a frequency band never alters your objective meters.

📚 Documentation Index · 🏗️ System Architecture · 🎛️ Audio DSP Chain · 🔊 Live Monitoring Internals

🗺️ Roadmap

✅ Shipped

• [x] Foundation: Multi-format WASM decoding, WebGL2 spectrograms, stereo metering, and phase correlation.
• [x] v0.2: 4-pane modular workspace, loudness / $F_0$ / onset timelines, and region WAV export.
• [x] v0.3: Built-in AutoEq headphone correction with global and workspace-scoped profiles.
• [x] v0.4: PAZ-inspired CQT spectrum analyzer and customizable peak hold.
• [x] v0.5: Anisotropic WebGL2 spectrogram rendering and interactive user onboarding.
• [x] v0.6: 8-file parallel playback, linked STFT zoom, and real-time multi-track spectrum overlays.
• [x] v0.6: Per-file loudness/feature extraction, shared Monitor deck, and selection mix bounce.
• [x] v0.6: Shared-region RMS level matching, same-rate residual diffing, and optional EQ baking in WAV exports.
• [x] v0.6: Reversible same-rate integer-delay alignment with optional polarity compensation.

🔭 Planned Next

• [ ] Harmonic & Pitch Intelligence: Chroma timelines and polyphonic melody estimation via PredominantPitchMelodia.
• [ ] Extended Spectral Descriptors: Optional feature lanes including real-time Spectral Centroid.
• [ ] Weighted Metering: Standard A / B / C frequency-weighted RMS measurements.
• [ ] Background Worker Pipeline: Off-thread Web Worker $F_0$ and onset computation for long-form recordings.

📋 Release History · 💡 Request a Feature

💻 Development & Contributing

Get a local development build running in seconds:

git clone https://github.com/Eps-Acoustic-Revolution-Lab/EAR-Audio-Preview.git
cd EAR-Audio-Preview
npm ci
npm run webpack-dev

1. Press F5 in VS Code to launch the Extension Development Host and open any audio file.
2. Run npm run ci to execute linting, tests, and bundle verification.
3. Run npm run vsix to package a standalone .vsix installer.

We warmly welcome bug reports, DSP discussions, and Pull Requests!

🐛 Open an Issue · 🛠️ Development & Release Guide · 🧪 Test Architecture & Guide

📜 Credits & License

Crafted with passion by Eps Acoustic Revolution Lab, originally inspired by and building upon the Custom Editor foundation of sukumo28/vscode-audio-preview.

Core DSP & Rendering Technologies:

Powered by Ooura · Essentia.js · twgl.js · wasm-audio-decoders · loudness-worklet · ebur128-wasm.

Released under the MIT License. All third-party libraries retain their respective licenses.