Forest Listeners: Tracking Wildlife with Smartphone Audio

You won’t find a jaguar on your phone, but you might hear one. A new app called Forest Listeners is turning the tedious work of cataloging jungle sounds into a game anyone can play from their couch.

It’s built on a simple, clever idea. Instead of relying on a few experts to parse millions of hours of rainforest audio, it asks the public to do it. The app drops you into a lush 3D simulation of the Amazon or Atlantic forest.

Your job is to listen for specific animal calls and click yes or no. That’s it.

Each click is a tiny piece of data. Collectively, those clicks are helping to train Perch, an AI model from Google DeepMind designed to automate biodiversity monitoring. The goal is to teach machines to hear the forest for the trees, to identify species by sound alone.

This matters because a forest’s health is written in its noise. The diversity and patterns of animal calls are a real-time diagnostic tool. But the recordings are endless, and AI models need vast, labeled datasets to learn.

For many species in the Brazilian rainforests, that training data simply doesn’t exist. Forest Listeners is an attempt to build it, one crowd-sourced click at a time.