Agentic WeeklyGet The Weekly Verdict, freeSubscribe free

Reported News

Hume's Audio Decision API answers plain-language questions about a recording with probabilities, no transcript

Hume's Audio Decision API, launched November 2, 2026, takes one audio file and questions written in plain language and returns probabilities instead of text. Hume's docs say it "returns no transcript, reply or explanation". Pricing is $0.47 per 1M input tokens; access starts from a waitlist.

By Agentic Weekly · Evidence: Reported · Editorial standards

Share on X
Changes reported
5
listed below, each from a source we read
We tested
Not tested yet
reported from sources; we did not run it ourselves
Gaps we flag
4
things we could not verify
Sources
4
4 primary, linked below

What changed

Changes the sources report, with the evidence status of each
#What the source saysStatus
01Hume's changelog dates the launch November 2, 2026: one endpoint, POST /v0/decisions/file, answers all your questions about one audio file in one request.Reported
02Each question uses one of four answer types: noul (yes/no), choice, score (a rating scale) or span (start and end times in the clip, for example where someone laughs).Reported
03Input is one MP3 or WAV file between 5 seconds and 5 minutes, plus an optional context field describing the recording; several recordings mean one request per file.Reported
04Billing is $0.47 per 1M input tokens (the audio plus the text of your questions); answers are not billed, and an organization gets a one-time 1M-token trial credit once access is granted.Reported
05There are Python (hume-audio-decisions) and TypeScript (@humeai/audio-decisions) SDKs, plus endpoints to estimate input tokens before a run and to list past runs.Reported

reported by the sourcemeasured or run by usnot verified

Primary source: Hume blog: Introducing Hume's Audio Decision API (Nov 2, 2026). We are the fast, independent index; the vendor’s own page is the source of record for this release.

Who it matters for

Teams building or testing voice agents, and anyone who scores call recordings. Hume lists evaluating text-to-speech output, testing voice agents, reviewing support calls and labeling speech corpora as the intended uses. Because the answer is a probability rather than text, your own code sets the threshold, for example flag a call when the chance the caller sounds frustrated passes 0.8.

It does not replace speech-to-text: it returns no transcript. Hume's own example pairs its answers with a separate transcription before handing both to an LLM that applies business rules. Live, real-time use is described by Hume as a longer-term opportunity; today's API takes recorded clips of 5 seconds to 5 minutes.

In the source’s words

“It returns no transcript, reply or explanation; your code does the reasoning with the numbers.”

“Raw audio goes in, an answer with a confidence score comes out, structured and ready for your code or your agent to act on.”

What we have not verified

  • We have not run the API; access is through a waitlist.
  • The 70.7% agreement with a human panel's average judgment (versus 56.5% for a single rater) is Hume's own initial evaluation; the test set is not published.
  • Speed and accuracy comparisons against other models are Hume's claims and were not checked.
  • Pricing is as listed on Hume's billing page on November 3, 2026.

This brief is Reported: written from the sources below, not from running the release ourselves.

Sources

Spotted an error? See our corrections policy or email the address on the contact page. Markdown version for agents: /news/hume-audio-decision-api.md

Want these briefs in your Google Top Stories? Add Agentic Weekly as a preferred source on Google.

Newsletter

The Weekly Verdict. Free, every Monday.

A free email every Monday: the week's few big AI releases, then 5 short use cases for the new models, each with a prompt you can paste.

Free. Double opt-in. One click to unsubscribe.