Audio Classification

Audio classification is the task of assigning a label or class to a given audio.

Example applications:

For more details about the audio-classification task, check out its dedicated page! You will find examples and related materials.

Recommended models

Explore all available models and find the one that suits you best here.

Using the API

from huggingface_hub import InferenceClient

client = InferenceClient(
    provider="hf-inference",
    api_key="hf_xxxxxxxxxxxxxxxxxxxxxxxx",
)

output = client.audio_classification("sample1.flac", model="firdhokk/speech-emotion-recognition-with-openai-whisper-large-v3")

API specification

Request

Headers
authorization string Authentication header in the form 'Bearer: hf_****' when hf_**** is a personal user access token with “Inference Providers” permission. You can generate one from your settings page.
Payload
inputs* string The input audio data as a base64-encoded string. If no parameters are provided, you can also provide the audio data as a raw bytes payload.
parameters object
        function_to_apply enum Possible values: sigmoid, softmax, none.
        top_k integer When specified, limits the output to the top K most probable classes.

Response

Body
(array) object[] Output is an array of objects.
        label string The predicted class label.
        score number The corresponding probability.
< > Update on GitHub