Hume AI

Free plan APIiPhoneMacWeb

What Hume AI is

Hume AI provides infrastructure for evaluating and improving voice AI. Its platform combines expression measurement, tests based on realistic interactions and human judgment to help teams assess voice models and agents. VoiceEQ creates targeted evaluation suites, tests systems under realistic conditions, diagnoses failures and helps validate changes. Its evaluation dimensions include reliability, emotion alignment, expressivity, naturalness, interruption handling and speaker verification. VoiceEQ supports WebRTC, WebSocket, SIP or live phone line, and batch audio connections, with TTS, speech-to-speech, ASR and speech-understanding modalities. Hume’s Expression APIs measure vocal and facial expression for live interactions and for analyzing recordings. The Audio Expression API offers 600+ dimensions across more than 50 languages; the Video Expression API measures 70+ expressions across faces in a frame. Hume also describes Prism Pipeline tools for cleaning, segmenting, transcribing, enriching and organizing audio, with deployment in a customer VPC on GKE, EKS or AKS. Its User Testing API uses people to test endpoints across languages, personas, scripts and edge cases, while its Human Feedback API returns human ratings in hours. Hume lists a free plan and paid plans from $3/mo. The service was founded in 2021 and is headquartered in New York City.

Who it suits

Hume AI suits teams building voice models or agents, and teams that need to evaluate voice systems in production. Its evaluation tools, expression measurement and human testing options address teams that need to assess voice interactions across languages and conditions.

What it does well

VoiceEQ tests systems under realistic conditions and helps diagnose failures.
Expression APIs measure vocal and facial expression.
Audio Expression API supports more than 50 languages.
Prism Pipeline can be deployed in a customer VPC on GKE, EKS or AKS.
Integrations include Agora, LiveKit, Twilio, Vapi and the Vercel AI SDK.
Offers a free plan and paid plans from $3/mo.

What to know first

Expression Measurement API access is in limited release; access can be requested from Hume.
Commercial use is not allowed.
Disabling EVI retention removes chat history and resume features.

Limits

The Expression Measurement API is in limited release, and access can be requested from Hume. For EVI, disabling retention removes chat history and resume features.

Getting started

Visit the Hume AI website.
Choose the free plan or a paid plan from $3/mo.
For Expression Measurement API access, request access from Hume.
Use the web, iOS or macOS service, or work through its API.
Consult the API SDK quickstarts for TypeScript, Python, .NET, or React or Next.js development.

Verdict

Choose Hume AI if you are building voice models or agents and need tools to measure, test or improve voice interactions. Its evaluation suite, expression APIs and human feedback options support several stages of that work. The stated restriction on commercial use makes it unsuitable for projects requiring commercial use.

Questions

Does Hume AI have a free plan?

Yes. Hume AI has a free plan, and paid plans start at $3/mo.

Which platforms does Hume AI support?

The listed platforms are web, iOS, macOS and API.

Can I use Hume AI commercially?

No. Commercial use is listed as not allowed.

Which languages does expression measurement support?

Hume states that expression measurement supports more than 50 languages.

Can I access the Expression Measurement API?

The API is in limited release, and access can be requested from Hume.

What integrations does Hume list?

Its developer page lists Agora, LiveKit, Twilio, Vapi and the Vercel AI SDK.

Facts from Hume AI's own pages

API access: The Expression Measurement API documentation says the API is in limited release and access can be requested from Hume. · dev.hume.ai, 7 Oct 2026
Audience: Hume says its products serve teams building voice models, fine-tuning voice agents, and evaluating production systems. · hume.ai, 7 Oct 2026
Audio pipeline: Prism Pipeline can clean, segment, transcribe, enrich, and organize audio, and is described as deployable in a customer VPC on GKE, EKS, or AKS. · hume.ai, 8 Oct 2026
Compliance: Hume's Voice API privacy documentation states that Hume AI is HIPAA compliant and allows users to request a BAA or DPA. · dev.hume.ai, 8 Oct 2026
Current focus: Hume describes its current offering as infrastructure for evaluating and improving voice AI. · hume.ai, 7 Oct 2026
Deployment: Hume says its Prism Pipeline can be deployed in a customer VPC on GKE, EKS, or AKS. · hume.ai, 7 Oct 2026
Developer tooling: The Expression Measurement API documentation lists Python and Node.js SDKs for its endpoints. · dev.hume.ai, 7 Oct 2026
Evaluation dimensions: The platform evaluates qualities including reliability, emotion alignment, expressivity, naturalness, interruption handling, and speaker verification. · hume.ai, 8 Oct 2026
Evaluation metrics: VoiceEQ lists measures including reliability, acoustic quality, expression understanding, interruption handling, speaker verification, and synthetic-speech detection. · hume.ai, 7 Oct 2026
Expression APIs: Hume’s Expression APIs measure vocal and facial expression to guide live interactions and analyze recordings for evaluation, training, and analysis. · hume.ai, 7 Oct 2026
Expression measurement: The Audio Expression API provides 600+ dimensions across 50+ languages, while the Video Expression API measures 70+ expressions across faces in a frame. · hume.ai, 8 Oct 2026
Founded: Hume AI's blog states that Alan Cowen founded Hume AI in 2021. · hume.ai, 8 Oct 2026
Headquarters: Hume lists its office at 51 Madison Ave, New York, NY 10010. · hume.ai, 7 Oct 2026
Human evaluation: The User Testing API uses real people to test endpoints across languages, personas, scripts, and edge cases; the Human Feedback API returns human ratings in hours. · hume.ai, 8 Oct 2026
Integrations: Hume's API documentation lists integrations including Vercel AI SDK, LiveKit, Pipecat, Vapi, Twilio, and Agora. · dev.hume.ai, 8 Oct 2026
Languages: Hume states that its expression measurement supports more than 50 languages. · hume.ai, 7 Oct 2026
Privacy: Hume’s privacy policy says it uses reasonable organizational, technical, and administrative measures to protect personal information. · hume.ai, 7 Oct 2026
Privacy controls: For EVI, users can disable retention of chat histories and voice recordings and opt out of anonymized interaction data being used for model training; these settings are opt-in by default and disabling retention removes chat history and resume features. · dev.hume.ai, 8 Oct 2026
SDKs: The API documentation navigation lists SDK quickstarts for TypeScript, Python, .NET, and React or Next.js development. · dev.hume.ai, 8 Oct 2026
Security and compliance: Hume’s Octave page lists SOC 2 Type II and HIPAA compliance for the product. · hume.ai, 7 Oct 2026
Support: Hume's API documentation directs users to its Discord community for support. · dev.hume.ai, 8 Oct 2026
Supported connection types: VoiceEQ lists WebRTC, WebSocket, SIP or live phone line, and batch audio as transports, and TTS, speech-to-speech, ASR, and speech understanding as modalities. · hume.ai, 8 Oct 2026
VoiceEQ: VoiceEQ combines automated measurement and human judgment to test voice AI under realistic conditions and identify areas for improvement. · hume.ai, 7 Oct 2026
Voice evaluation: VoiceEQ generates targeted evaluation suites, tests systems under realistic conditions, diagnoses failures, and helps validate improvements. · hume.ai, 8 Oct 2026