The audio editor is a labeling tool for customers interested in building machine learning models that can detect, classify, and transcribe audio. The current version of our audio editor supports applying all classification types to audio files.

Labelbox currently accepts MP3, WAV, and M4A audio files. When you attach an audio dataset to a project, the Labelbox Editor will automatically adjust the Editor interface for audio labeling.

📘

Supported file formats

• MP3
• WAV
• M4A

Supported annotation types for the Audio editor

Below are examples of how each annotation type will be formatted in the export file.

👍

Supported annotation types

(Classifications can currently be applied at the asset level only.)

• Radio classification
• Checklist classification
• Free-form text classification
• Dropdown classification

How the Audio editor works

Playing the audio file

Audio files display in the Labelbox editor with a waveform for visualization and basic player controls. Use the play/pause button to begin audio playback, and press the back/forward 10-second buttons to quickly replay or advance playback. You can also click anywhere on the waveform to instantly move the playhead to your desired location.

Classifying the audio file

The classification questions added to the labeling project in project setup will appear in the left-hand panel. Labelers may use their mouse or the specified keyboard shortcuts to activate various questions and apply the relevant values.

Submitting the audio label

Once the labeling task is completed, the labeler may click Submit (or press e) to submit the completed label and advance to the next asset.