Second Listen
Two independent models, BirdNET-Analyzer and Google DeepMind's Perch, listen to the same audio and don't always agree on what they hear. Upload your own recording below and watch both models analyze it live, or browse how they compare across a set of public Xeno-canto recordings further down. Personal portfolio project, not affiliated with American Bird Conservancy or the BirdsPlus Index.
Analyze your own recording
Upload an mp3, wav, m4a, ogg, or flac file, or record one right here, and both models run on it directly, the same BirdNET and Perch passes used on the Xeno-canto dataset below, just live instead of pre-computed. Nothing here is pre-loaded: whatever you upload gets analyzed on the spot, and results aren't mixed into the species chart or map, they only ever show up here.
Sharing this can make BirdNET's guesses more accurate, since it narrows down which species are realistic for that place and time of year.
Off by default: boosting volume isn't free, on one quiet test recording it changed which species the model guessed, so it's worth trying both ways rather than assuming enhancement can only help.
Example recordings from around the world
Each marker represents a recording from the public Xeno-canto dataset, not your own uploads. Each marker sits at that recording's actual location and is sized by how many detections it produced. Click one, then "View sonogram & play" to see exactly where each model detected a species and hear that moment for yourself.