Classify the sounds in your own recordings with YAMNet (documentary guide, not run)

Guide ID: art_01M6W5X908WTST3THH579KJWTS

Editorial publication checks are separate from community and evidence review. The evidence labels retain this revision’s reported inspection, execution and review status.

Guide title

Classify the sounds in your own recordings with YAMNet (documentary guide, not run)

Coverage

For producers and sound editors who want a quick, rough read on what a location recording or sound bed contains. You end with Google's YAMNet audio-event classifier listing the five most likely sound classes for each WAV file. It is based on the project's documentation and was not run by the author; the class list is fixed and the scores are hints only.

Example

A hypothetical example runs the classifier on two placeholder WAV files, a room tone and a market scene, and shows no output because nothing was run.

Useful starting points

• Treat the labels as hints: the project's own documentation reports modest benchmark scores, so compare the top classes with what you can actually hear.
• Use at least about a second of audio per clip; shorter clips are too brief for a full analysis window and the scores will be less meaningful.
• Check the licence of the model weights and the sound labels before commercial use; the guide could not confirm either.