Skip to content
Hume Decision pricing

Hume Audio Decision API cost calculator (token → USD)

Approximate input tokens and USD from clip length, questions, and windowing — at Hume’s published $0.47 per 1M input tokens. Cite: official billing. This page does not call the Estimate API or upload audio.

Context size
Windowing

Estimated input tokens

26,288tokensapprox

Estimated cost (USD)

$0.012355@$0.47 / 1M input tokens

Trial runs left (1M tokens)

38

Breakdown

audio ~26,120 · questions ~168 · context ~0 · 1 call(s)

Approximate — calibrated from Hume’s public examples, not a live Estimate API response.

  • Duration (seconds). Allowed range 5–300 seconds (5,000–300,000 ms), same as estimate/decision limits.
  • Questions. How many decision questions you attach (1–20). Official examples often use 6.
  • Context size. Extra prompt context beyond the audio. Short ≈ one sentence; long ≈ a paragraph.
  • Windowing. Whole clip once, or many shorter calls (30s chunks, overlapping 6s, or live 5s). Overlap multiplies cost.

How we estimate tokens

We use a transparent client-side model calibrated to Hume’s published rounded examples (about ±10%). Audio length dominates; each question and context size add a small fixed amount; windowing repeats the call. Answers are not billed. We never send your audio or call Hume’s Estimate endpoint from this page.

Price we use

$0.47 per 1M input tokens, from Hume’s Audio Decisions billing docs (checked Nov 3, 2026). 4xx rejects and the Estimate endpoint itself are not billed. Org trial is 1M input tokens once access is granted.

FAQ

Is this the official Estimate API?

No. This is an approximate front-end calculator. For exact counts use Hume’s Estimate endpoint in your own stack.

What does $0.47 / 1M cover?

Input tokens only. Answers are not billed.

Why can windowing cost more for the same audio length?

Overlapping or chunked windows are separate billable calls; questions/context can apply per call.