Inference

Say a command. Keep the audio.

Audio is decoded and classified on this device. Nothing is uploaded. The model expects one second of 16 kHz audio and recognizes ten commands plus background and unknown.

A

Select input type

Choose an audio source

Input source
Listen continuously After 3–2–1, say commands with a short pause between them. Stop when finished.

Waiting for audio.

B

Model output

Top prediction

— Add audio to begin

Spectrogram

log magnitude

Waveform

1 second · 16 kHz