Hume launches Audio Decisions API: ask any question about audio, get probabilistic judgments
rohanpaul_ai · x · 2026-11-03
Hume AI launched its new Audio Decisions API, positioned as a decision model for voice AI: developers define the questions, and the model returns probabilistic judgments straight from audio. Key features:
- One model, any question: ask objective, subjective, or perceptual questions about the same recording — e.g., "Does the patient sound anxious?", "How many speakers are audible?";
- Chunked evaluation: audio is answered in 5-second segments, suited to QA of calls like telehealth consultations;
- Built for scale: re-run evaluations across collections to surface training candidates, isolate eval failures, and flag cases for escalation; claims 180ms per file scanned.
The product is currently waitlisted, with API docs available.
Related event: Hume AI launches Audio Decision API, an audio-native decision model(6 posts)→
More from Multimodal
- Cosmos 3 world model follows camera controls from a single image, recipe released — NVIDIAAI · 2026-11-03
- NVIDIA Cosmos Simulation Recipes Turn Daytime Drives Into Rainy Nights With Trajectories Intact — NVIDIAAI · 2026-11-03
- NVIDIA Cosmos recipes turn depth video into RGB scenes for robot and AV training — NVIDIAAI · 2026-11-03
- Suno user begs for a button that writes prompts so they never have to — gleech · 2026-11-03
- Seedance 2.5 demo: a 30-second swamp creature short film made in Dreamina — LudovicCreator · 2026-11-03
- Turning his photo library into drum samples with Gemini 4 — alexanderchen · 2026-11-03