Blog

Earth Species Project welcomes Steven VanRoekel as CEO. In this Q&A, Steve shares his vision for decoding animal communication with AI, his personal connection to wildlife, and what the next phase of growth looks like for ESP.

Lightweight biologgers and ESP’s machine learning tools reveal the hidden vocal world of wild carrion crows, capturing over 127,000 calls from quiet murmurs to long-distance signals.

The first global survey on AI and animal communication reveals striking consensus: people worldwide recognize animals’ complex inner lives and demand technology that fosters connection, not exploitation.

To better understand what matters for building a generalizable bioacoustic encoder, we tested 19 models across 26 datasets and a new evaluation benchmark. Our main finding is that a two-stage training approach—self-supervised pre-training followed by supervised post-training, both on a mix of bioacoustic and general audio—delivers the strongest performance.

Explore NatureLM-audio, our open-source model for bioacoustics. Try the new demo to analyze animal sounds, identify species, and support research.

FrogID, the world’s largest frog-focused citizen science project, tested Earth Species Project’s open-source NatureLM-audio model to help automate the validation of frog vocalizations submitted by the general public. Without fine-tuning, the model excelled at distinguishing frogs from other sounds, potentially saving the team 300+ hours of manual work annually. It also showed strong accuracy on…

Earth Species Project and Raincoast Conservation Foundation have partnered on a new research project to decode killer whale communication with AI.

Explore Earth Species Project’s 2024 breakthroughs in AI and animal communication—and a bold new strategy to help life on Earth thrive.

Ethologists and bioacousticians are constantly battling noise, from the sound of wind or rain or even the animals’ own movements, drowning out the vocalizations they’re trying to analyze. We introduce a new method, Biodenoising, to address these challenges by denoising animal vocalizations without requiring clean data.