Hume AI launches Audio Decisions API for native speech evaluation Other

Hume AI released Audio Decisions, an API endpoint that analyzes voice recordings directly using native audio modeling to answer questions with structured judgments and confidence scores. The API is available now in developer preview.

Unlike traditional voice processing pipelines that transcribe speech to text before querying an LLM, Audio Decisions processes vocal acoustics natively. The system evaluates prosody, pauses, pitch variation, and emotional cadence to detect customer sentiment, conversational hesitation, or tension.

The API features speaker separation to isolate specific participants while ignoring background noise. Hume provides two tiers: a standard evaluation tier at $0.009 per minute of audio and an advanced tier with granular emotional analysis at $0.018 per minute.

What it is

• Native audio evaluation. Ingests audio directly without intermediate speech-to-text conversion to evaluate vocal inflection and tone.
• Structured output. Emits schema-validated answers alongside probabilistic confidence scores for automated decision logic.
• Speaker diarization. Isolates designated speakers and filters background chatter and non-speech acoustic artifacts.
• Customer operations. Targeted at compliance checks, customer service QA scoring, and conversational user experience audits.

Price

Sources

• Hume AI: Introducing Hume Audio Decision API
• Hume AI: Decision Model API product page