WildEchoLiving sound archive

On-device recognition · microphone audio is never uploaded

Translucent bioluminescent bird sound specimen

WILDECHO COLLECTION · 001

Hear the shape
of life.

Point toward nature and listen. Species are identified on your device, and the sound never leaves your browser.

Loading BirdNET model · 0%

LIVE SOUND SPECIMEN

REAL BIRD SOUNDS · LISTEN & IDENTIFY

Wikimedia Commons · CC BY-SA 3.0 / Public Domain / CC BY-SA 4.0

Record for 8–15 seconds
Move closer and reduce human speech

正在拆解声纹…

能量 · 频谱 · 节律 · 物种模型

A field instrument for living sound

From a call
to evidence.

WildEcho turns a short recording into a local, inspectable observation. It does not turn uncertainty into certainty: every result keeps its confidence, alternatives, and acoustic evidence attached.

01 · LISTEN

Capture the habitat

Record 8–15 seconds in the field or import an existing clip. The browser keeps a 48 kHz mono signal on your device.

02 · IDENTIFY

Compare candidates

BirdNET evaluates overlapping sound windows and returns ranked species candidates. Results below the evidence threshold remain unclaimed.

03 · INSPECT

Keep the evidence

Duration, level, dominant frequency band, source and model threshold stay visible beside the identification.

Privacy boundary

The sound stays here.

Recognition runs inside the browser. Microphone audio and imported recordings are not sent to WildEcho servers.

Scientific restraint

A candidate, not a verdict.

Species scores are machine estimates. Field conditions, distance and background noise matter, so WildEcho preserves uncertainty instead of hiding it.

Open foundations, visible method.

BirdNETTensorFlow.js48 kHz PCMOn-device inference