Hume

What is Hume?

Hume is an emotionally aware AI voice platform made by Hume AI that detects, understands, and expresses emotion in voice interactions. The platform provides open-source models, datasets, and evaluation APIs designed to embed emotional intelligence into voice AI systems for applications ranging from customer service to gaming and healthcare.

The core offering centers on three proprietary models: TADA (an open-source LLM text-to-speech system that streams text and audio together to reduce hallucinations and latency), Octave (a closed-source LLM text-to-speech system with voice design, voice modulation, voice cloning, and voice conversion capabilities), and EVI (a closed-source speech-to-speech system featuring interruptibility, back-channeling, expressive instruction following, and external LLM compatibility). These models are trained on multimodal emotional intelligence research spanning 50+ languages, 48+ distinct emotions, and 600+ voice descriptors. The company also provides curated speech datasets annotated for conversational audio dynamics, emotional reproduction, multilingual native speaker recordings, voice realism metrics, and domain-specific or task-specific applications in healthcare, finance, scheduling, and customer support workflows.

Hume's Human Feedback API enables developers to run human evaluation studies with science-backed survey templates, drawing from a worldwide pool of vetted raters to assess model preference data. The platform offers fast turnaround times (hours rather than weeks) and combines listenability, audio quality, and smoothness metrics for comprehensive evaluation. A RESTful API is available for programmatic study management, though it is noted as coming soon on the website. The tool operates on a freemium model, allowing teams to access documentation, join the developer community via Discord, and explore open-source resources, with paid plans for production use and advanced features.

Organizations using Hume include dev