Hume AI

O que é Hume AI?

Hume AI provides software and APIs for voice evaluation and expression measurement. VoiceEQ helps teams define tests, examine realistic conversations, combine automated measures with human feedback, and investigate failures. Expression Measurement APIs return structured scores from recorded or streamed media. Hume's legacy Octave TTS and EVI voice APIs remain supported until their announced shutdown on November 13, 2026.

O que Hume AI pode fazer?

Voice evaluation workflows

Generate use-case-specific test scenarios and success criteria, compare systems, and use findings to guide another round of testing.

Expression measurements

Measure vocal expressions and speaking qualities in speech, or visible facial expressions and actions in images and video frames.

Recorded and live media

Send existing media through upload endpoints or receive ongoing measurements from WebSocket audio and video streams.

Developer integration

Use Python or Node.js SDKs and associate measurements with speech segments, detected faces, and recorded request runs.

Legacy voice generation

Octave converts text to expressive speech; EVI supports real-time voice interaction. Both are subject to the announced Voice API sunset.

Casos de uso

“Test an internal voice application”

Evaluate timing, expression, and interaction quality when comparing options for your own application.

“Annotate audio for review”

Connect expression scores to speech segments to identify moments for closer human examination.

“Plan a voice API transition”

Move existing TTS or EVI integrations and export relevant account data ahead of the announced shutdown.

Perguntas frequentes

When do TTS and EVI end?

Access ends November 13, 2026 at 00:01 EST. Hume says both remain fully supported until then. The notice in its Voice API documentation schedules permanent deletion of Voice API account data after that date.

Can anyone immediately use the new APIs?

Expression Measurement is in limited release and requires an access request. VoiceEQ also provides a request-access process.

How is Expression Measurement usage counted?

Audio is billed by its full duration, including silence. Each image or video frame counts as one third of a second. Each billable run has a minimum of ten seconds.

What data and usage permissions apply?

Provide media you have permission to process. Standard terms limit usage to developing and supporting your own applications, exclude regulated medical and other prohibited data, and grant Hume rights to improve and train its services, subject to the applicable order.