Skip to main content
POST
Zero-shot audio classification

Body

application/json

Zero-shot classification request for audio inputs using candidate labels.

audios
number[][]
required

Audio input(s) as list(s) of floats (waveform arrays). Required.

candidate_labels
string[]
required

List of candidate labels to classify the audio against. Required.

hypothesis_template
string
default:This is a sound of {}

Template used to format each candidate label into a hypothesis for scoring.

Response

200 - application/json

Zero-shot audio classification results

label
string

Predicted label name.

score
number

Confidence score (probability-like) for the label.