For Hume AIAI Platform

Hume Octave Tts

Hume AI (EVI / Octave) · Hume AI

Empathic Voice AI — Hume AI

Hume AI evals — Octave TTS (relift v3 InfraRed)

About Hume AI

Hume AI builds empathic voice technology — the Empathic Voice Interface (EVI), an expression-measurement API across face, voice, and language, and the Octave text-to-speech model — for emotionally aware, responsibly governed voice experiences.

Employees

~60

Industry

Voice AI

Headquarters

New York, NY

Website

hume.ai

Sample tests· showing 3 of 9

#	Input	Expected behavior	Check
01	Operator wants Octave to read a refund apology 'calmly and warmly, with a slight pause before the apology'.	Put the acting/delivery guidance in the utterance's description field and the words to be spoken in text. Octave interprets description as performance direction, not as content to read aloud.	Pass / FailAi Platformhigh
02	Operator can either pick an existing Octave voice by id/name or describe a new voice to design; they pass both a voice id and a voice-design prompt.	Choose one path per request: reference a saved voice (by id/name) for consistency, or use a voice-design description to synthesize a new voice. Do not send a conflicting saved-voice id and a contradictory design prompt in the same request.	Pass / FailAi Platformmedium
03	Text contains '$1,250.50 due on 03/04' and the operator wants it spoken naturally and unambiguously.	Rely on Octave's text normalization for currency/dates, and where ambiguity matters (03/04 day-vs-month), pre-normalize the text to an explicit form ('March 4th') so the spoken output is unambiguous. Verify the rendered speech matches intent.	Pass / FailAi Platformlow
Use this eval

How this eval is graded

Grade against expected.ideal_behavior and expected.rubric. Per-criterion pass requires mean >= 4.0 and no criterion below 3. Emotion-expression scores are probabilistic perceived-expression signals, not ground-truth affect.