Changelog

The Expression Measurement API is now available in limited release. It measures emotional expression in speech and in faces, from uploaded audio and images or live audio and video streams. Contact our team to request access, then start with the overview.

• Audio endpoints. Audio upload measures an audio file in one request, and audio realtime measures audio as you stream it over a WebSocket. Each measurement scores the expressions and voice_attributes of a window of speech.
• Video endpoints. Image upload measures JPEG images in one request, and video realtime measures video as you stream it, one JPEG frame at a time. Each face detected gets a bounding box, and the largest faces in each image, 2 by default, also get scores for their expressions and descriptions.
• Run endpoints. Runs return the status, configuration, reported problems, and billable seconds of each upload request and realtime session.
• SDKs. Python (hume-expression-measurement) and Node.js (@humeai/expression-measurement-node) SDKs cover every endpoint. See the Python quickstart and the Node.js quickstart.