Hume AI

Emotionally intelligent voice AI with an empathic interface and expression measurement, built for developers.

Hand-reviewed by the Toolglade team·Updated Aug 2026

audio#voice-ai#emotional-ai#developer-platform

Free plan Free trial Claimed API Teams

Toolglade’s take

Hume AI stands out for grounding its voice models in genuine emotion-science research, making it a strong pick for teams who need empathic, expression-aware voice rather than just natural-sounding TTS. Usage-based, per-minute pricing keeps entry cheap but scales with volume.

About Hume AI

Hume AI is an emotionally intelligent voice AI platform whose Empathic Voice Interface (EVI), Octave TTS, and Expression Measurement API let developers build conversational agents that detect and respond to emotion in real time. It is grounded in emotion-science research covering 48+ emotions and 50+ languages, and sold through usage-based, per-minute and per-character APIs. A free tier and low-cost paid plans make it accessible to individual developers as well as enterprise voice teams.

Hume AI is a New York-based AI research company building tools that give voice and conversational AI emotional intelligence. Its flagship Empathic Voice Interface (EVI) is a speech-to-speech model that measures vocal expression and adapts its own tone in real time, while the Octave text-to-speech models generate expressive, controllable voices and the Expression Measurement API analyzes emotion across 48+ categories and 50+ languages. The platform is aimed squarely at developers and voice AI teams, delivered through APIs and SDKs with usage-based pricing. More recently Hume has expanded into evaluation and data tooling, including the Kairos simulation platform, a Human Feedback API, and public Voice AI leaderboards, positioning itself as both a builder and a measurement layer for human-quality voice AI.

TL;DR

Hume AI is an emotionally intelligent voice AI platform whose Empathic Voice Interface (EVI), Octave text-to-speech, and Expression Measurement API let developers build agents that detect and respond to human emotion in real time. It is grounded in decades of emotion-science research spanning 48+ emotions and 50+ languages. Pricing is freemium and usage-based, with a free tier and paid plans from $3/month up to custom Enterprise. It targets developers and voice AI teams rather than end consumers.

Company overview

Hume AI, Inc. is a New York-based artificial intelligence research company focused on measuring and integrating human emotional expression into technology. It was founded by Alan Cowen, a former Google DeepMind researcher and computational emotion scientist, and its work draws on large-scale research into how emotion is expressed across voice, face, and language. The company positions itself as building 'AI with emotional intelligence to create technology that truly understands humanity.'

Beyond its voice models, Hume has expanded into an evaluation and data layer for voice AI, offering the Kairos simulation platform, a Human Feedback API backed by real human raters, custom data-collection solutions, and public leaderboards such as Real World VoiceEQ Bench that rank voice AI models by human judgment.

Product features

Hume's core products are the Empathic Voice Interface (EVI), a speech-to-speech model that measures vocal expression and modulates its own responses; Octave (versions 1 and 2), expressive and controllable text-to-speech engines with unlimited voice cloning; and the Expression Measurement API, which analyzes emotion in real time across 48+ categories, 50+ languages, and 600+ output descriptors. EVI supports bringing external LLMs so teams can pair Hume's empathic voice layer with their own reasoning model.

The platform is API- and SDK-first (TypeScript and Python), with a web app, playground, and documentation at dev.hume.ai. Higher tiers add more concurrent connections, higher request rates, team seats, and enterprise-grade compliance (SOC 2 Type II, GDPR, HIPAA). Newer offerings, Kairos and the Human Feedback API, let teams simulate conversations at scale and collect human ratings to close the model-evaluation loop quickly.

Target market

Hume AI targets software developers, product teams, and voice AI companies that need emotionally aware speech and conversational capabilities, including builders of AI companions and characters, customer experience and support analytics teams, healthcare and wellbeing apps, and research groups. Its usage-based tiers scale from individual developers on the free or Starter plans up to enterprises requiring high volume, compliance, and dedicated support.

Buyer personas

End users

Software developers and ML engineers who integrate EVI, Octave TTS, or the Expression Measurement API into voice apps, agents, and analytics pipelines.

Buyers

Engineering leads, product managers, and CTOs at startups and enterprises choosing a voice AI vendor and approving usage-based or enterprise contracts.

Key influencers

AI researchers, conversation designers, UX teams, and compliance stakeholders who evaluate emotional accuracy, voice quality, and data-privacy posture.

Ideal customer profile

A product or platform team building voice-driven or conversational AI where emotional intelligence and expressive, empathic voice materially improve the user experience, and that is comfortable integrating APIs on usage-based pricing.

Funding & performance

Hume AI raised a $50 million Series B round announced in March 2024, led by EQT Ventures with participation from Union Square Ventures, Nat Friedman & Daniel Gross, Metaplanet, Northwell Holdings, Comcast Ventures, and LG Technology Ventures, bringing total funding to roughly $60+ million. Any figures beyond this should be treated as not publicly disclosed.

Pros & cons

Pros

Grounded in peer-reviewed emotion-science research
Real-time emotion detection across 48+ categories and 50+ languages
Expressive, controllable voice output via Octave TTS
Low-cost entry tier plus a genuinely usable free plan
Supports external LLMs so you can bring your own model
Strong developer tooling with SDKs, docs, and a playground

Cons

No self-hosted or on-premise option
Usage-based pricing can grow quickly at high volume
No mobile app or browser extension for end users
Focused on developers, so non-technical users need engineering help
Emotion measurement is probabilistic, not a guaranteed ground truth

Pricing plans

Free

$0 / month

10,000 TTS characters (~10 min)
5 EVI minutes included
15 requests per minute
1 concurrent connection
Unlimited voice cloning (create and use)
Discord support

Starter

$3 / month

30,000 TTS characters (~30 min)
40 EVI minutes ($0.07/min after)
15 requests per minute
5 concurrent connections
Unlimited voice cloning
Discord support

Creator

$14 / month

140,000 TTS characters (~140 min)
$0.15 per 1,000 extra characters
200 EVI minutes ($0.07/min after)
75 requests per minute
5 concurrent connections
Unlimited voice cloning

Pro

$70 / month

1,000,000 TTS characters (~1,000 min)
$0.12 per 1,000 extra characters
1,200 EVI minutes ($0.06/min after)
75 requests per minute
10 concurrent connections

Scale

$200 / month

3,300,000 TTS characters (~3,300 min)
$0.10 per 1,000 extra characters
5,000 EVI minutes ($0.05/min after)
150 requests per minute
20 concurrent connections
3 team seats

Business

$500 / month

10,000,000 TTS characters (~10,000 min)
$0.05 per 1,000 extra characters
12,500 EVI minutes ($0.04/min after)
225 requests per minute
30 concurrent connections
5 team seats

Enterprise

Custom

Custom usage volumes
Custom rate limits and connections
Unlimited team seats
Voice cloning via API
SOC 2 Type II, GDPR, HIPAA compliance
Slack support

Key features

API

Team collaboration

Multi-language

Integrations

API/SDK

Input types

voice, text

Output types

voice

Best For

Voice AI developers, Empathic conversational agents, Emotion analytics, Expressive text-to-speech

Compare key features

View all alternatives →

Feature

Hume AI

Cartesia

Vapi

Pricing

Freemium

Free plan

Yes

Free trial

Yes

API

Yes

Team support

Yes

Frequently asked questions

What is Hume AI and its Empathic Voice Interface (EVI)?+

Hume AI is an AI research company building emotionally intelligent voice technology. Its Empathic Voice Interface (EVI) is a speech-to-speech conversational AI that measures the tone and emotion in a user's voice and adapts its own responses in real time, alongside products like Octave text-to-speech and an Expression Measurement API.

How much does Hume AI cost?+

Hume AI is freemium with usage-based pricing. There is a free plan at $0/month, and paid plans start at Starter for $3/month, then Creator $14, Pro $70, Scale $200, and Business $500 per month, with each tier including more TTS characters and EVI minutes. Overages are billed per 1,000 characters or per minute, and Enterprise is custom-priced.

How does Hume AI compare to ElevenLabs?+

ElevenLabs focuses primarily on high-quality, natural-sounding text-to-speech and voice cloning, while Hume AI centers on emotional intelligence, its EVI can detect vocal emotion and adjust its tone, and it offers a dedicated Expression Measurement API. Choose Hume if empathic, emotion-aware voice interaction is the priority; choose ElevenLabs if you mainly need broad, polished TTS voices.

Is Hume AI meant for developers?+

Yes. Hume AI is delivered as APIs and SDKs (TypeScript and Python) with documentation at dev.hume.ai, a playground, and usage-based plans, making it primarily a developer platform for building voice and emotion features into applications rather than a finished consumer product.

How accurate is Hume AI's emotion detection?+

Hume's models are trained on large, research-grade datasets and measure expression across 48+ emotion categories and 50+ languages with 600+ output metrics. The output is probabilistic, it estimates the likely emotional expression in voice, face, or language rather than reading true internal feelings, so it is best used as a strong signal rather than an absolute ground truth.