Audio inspection for speech, ML, and signal engineering workflows inside VS Code.

Waveform · Spectrogram · Multi-channel tracks · Selection analysis · Raw PCM · Kaldi WAV Ark · Remote SSH

English | 简体中文 | 日本語

AudioLens turns VS Code into a practical audio viewer for audio engineers, speech engineers, and ML practitioners. Open common audio files, raw PCM dumps, or Kaldi WAV Ark entries directly beside manifests, transcripts, logs, scripts, and model outputs.

It focuses on the daily workflow that generic audio players miss: inspect waveforms and spectrograms, review multi-channel files, open audio paths from text, decode raw PCM with explicit parameters, inspect headers, and analyze selected regions without leaving the workspace.

Install from VS Code Marketplace · Install from Open VSX · Download VSIX

Recommended: install from the Visual Studio Marketplace

https://marketplace.visualstudio.com/items?itemName=simzhou.audiolens

Alternative: install from Open VSX

https://open-vsx.org/extension/simzhou/audiolens

Command line

code --install-extension simzhou.audiolens

Offline VSIX

code --install-extension dist/audiolens-1.4.2.vsix

• Method 1: Ctrl-click a wav.ark:offset path. Requires Kaldi Reader: GitHub, VS Code Marketplace, or Open VSX.
• Method 2: Open an .ark file and enter the offset manually. No additional extension is required.

AudioLens uses the browser audio stack for common encoded formats and the extension host to read files from the VS Code workspace.

Multi-channel files are shown as separate channel tracks. Each track has a compact left control strip and a full-width analysis area.

• Mute disables playback for that channel.
• Solo plays that channel and silences the other channels.
• The track view selector switches a channel between waveform, spectrogram, and combined view.
• Selecting a track makes it the active channel for selection analysis.

The waveform color is consistent across channels so the selected channel does not visually distort track comparison. Adjacent tracks are drawn as a compact stack with shared borders, while the selected track keeps a rounded focus outline for quick orientation.

For .pcm and .raw files, AudioLens asks for PCM parameters before decoding:

• sample rate
• channel count
• encoding, such as Signed 16-bit PCM, Unsigned 8-bit PCM, 32-bit float, or 64-bit float
• byte order, with 8-bit encodings automatically using no endian setting
• start offset in bytes

The current PCM parameters can be saved as defaults for later PCM files. AudioLens does not guess PCM parameters from the file name, because raw PCM does not contain reliable metadata.

WAV files can also be reopened as PCM from the top bar. This is a one-time operation for the current file and is useful when inspecting raw audio data, non-standard headers, or offset-sensitive test files.

Run AudioLens: Open Kaldi WAV Ark Entry from the Command Palette and enter a wav.ark:offset location. If you open an .ark file directly, AudioLens asks for the offset before reading.

AudioLens only supports ark entries whose audio payload starts with a WAV RIFF/WAVE header. It uses the WAV header size to read the selected entry and does not scan or load the whole ark file.

AudioLens can detect audio paths in ordinary text files and open them directly with the AudioLens editor. Hover an audio path and click Open in AudioLens, or place the cursor on a path and use the status-bar action or AudioLens: Open Audio Path at Cursor. It supports absolute paths and relative paths resolved from the current text file, workspace folders, and optional configured base directories.

Run AudioLens: Toggle "Open in AudioLens" from the Command Palette to turn this feature on or off. It is enabled by default and avoids generating inline links for the whole document, so large JSON, log, and dataset files stay responsive.

Kaldi *.ark:offset links are intentionally left to Kaldi Reader.

Use the document icon in the top bar to inspect structured header fields without leaving VS Code. AudioLens lists fields in file order and uses byte offsets for chunk-based formats, or bit ranges for packed headers such as ADTS AAC and MPEG audio frames.

For WAV files, the inspector highlights whether the file uses the standard 44-byte PCM header or contains extended chunks such as fmt extensions and LIST metadata. Audio payload rows identify the data region without dumping raw sample bytes.

Drag across any waveform or spectrogram to create a time selection. AudioLens can play the selected range and calculate metrics for the active channel.

Current analysis includes:

• start time, end time, and duration
• RMS level and peak level
• dominant frequency
• crest factor
• clipping ratio
• noise floor estimate
• spectral centroid
• zero-crossing rate
• frequency-band distribution

Tooltips next to the metrics describe how each value is calculated and when it is useful.

AudioLens includes practical spectrogram controls for speech and signal inspection:

• algorithms: Frequency, Reassignment, Pitch (EAC)
• FFT sizes from 8 to 32768
• window functions: Rectangular, Bartlett, Hamming, Hann, Blackman, Blackman-Harris, Welch, and Gaussian variants
• zero padding factors from 1 to 128
• frequency scales: Linear, Log, Mel, Bark, ERB
• configurable display-only frequency range, with an optional Nyquist-following maximum
• palettes: Rose, Classic, Grayscale, Inverse Grayscale
• configurable dB brightness range and auto brightness

The settings menu in the top-right corner keeps these controls close to the spectrogram view, including display-only frequency range limits and Nyquist-following maximum frequency.

Spectrogram work runs in a worker so expensive analysis does not block Webview interactions.

After opening audio, the active spectrogram or waveform is keyboard-ready, so Space can play or pause immediately.

AudioLens follows the VS Code display language by default. You can override the Webview language with the audiolens.language setting or by running AudioLens: Switch Language from the Command Palette.

Supported languages:

Simplified Chinese, Traditional Chinese, English, Japanese, Korean, French, German, Russian, Spanish, Italian, Portuguese, Indonesian, Norwegian, Dutch, Polish, Turkish, and Vietnamese.

New interface strings fall back to English until a locale has a complete translation.

AudioLens is declared as a workspace extension. In a Remote SSH window, the extension host runs in the remote workspace, reads audio files from the remote file system, and streams the data to the local Webview for playback and visualization.

Use the top-bar download button when you want to save the current remote audio file. VS Code may open the save dialog on the remote side first; choose the local location option in that dialog when saving to your machine.

AudioLens does not upload audio files to any third-party service. Audio content is read by the VS Code extension host and analyzed inside the VS Code Webview and worker runtime.

npm install
npm run build
npm run typecheck
npm run rust:test
npm run package

Press F5 in VS Code and choose the AudioLens extension launch configuration. Then open a supported audio file in the Extension Development Host.

SimZhou: https://simzhou.com/en/about/

If AudioLens helps with your speech, audio, or signal engineering workflow, you are welcome to support its ongoing development.

Support AudioLens on Ko-fi: https://ko-fi.com/simzhou

Copyright (c) 2026 SimZhou. All rights reserved.