Hume AI launches Audio Decision API, an audio-native decision model

On November 3, Hume AI released the Audio Decision API, positioned as "the decision model for voice AI." Developers can ask any objective, subjective, or perceptual question about a recording in natural language—such as how many people are speaking, whether anyone is laughing, or whether a caller sounds frustrated—and the model returns probabilistic judgments based directly on the audio itself.

Confirmed

• Hume AI officially launched the Audio Decision API (some posts call it Audio Decisions API / Decision API), built on what it calls the world's first audio-native decision model; a single call replaces the 10+ models that previously had to be chained together.
• The model was trained on 10M+ human auditory judgments, so its answers align closely with how real human listeners would judge.
• The core selling point: even a perfect transcription loses tone information (hesitation, frustrated intonation, interruptions, etc.), while analyzing audio directly preserves these signals.
• @cneuralnetwork gave a typical example: in a customer service call, a user might say "sure, fine" while already annoyed, and text transcription drops that signal—voice agents often fail to pick it up. The Decision API returns probability-based tone and emotion judgments that can be used to quantify voice agent performance.

Why it matters

• Traditional voice AI pipelines center on transcription, and paralinguistic information such as tone and emotion is lost at that step; Hume's approach moves decision-making up to the audio-native level.
• Compressing multi-model chains into a single call reduces development complexity and cost for voice applications (especially customer service and voice agent scenarios), and provides a new tool for emotion awareness and performance measurement in voice agents.

2026-11-03 ~ 2026-11-03 · 6 related posts

• Hume Launches Audio Decision API Answering Voice Questions in One Call — EXM7777 · 2026-11-03
• [source] Hume AI Launches Decision API: Analyzes Audio Directly to Score Voice Agent Tone and Emotion — cneuralnetwork · 2026-11-03
• Hume launches Audio Decision API to replace 10+ audio models with one call — alifcoder · 2026-11-03
• [source] Hume launches first audio-native decision model that answers structured questions straight from recordings — testingcatalog · 2026-11-03

2 near-duplicate retellings: Scobleizer · rohanpaul_ai