Hume

Hume offers AI technology that's been trained to improve emotional intelligence in a variety of settings. The company's platform offers a variety of tools, including the Empathic Voice Interface (EVI), Expression Measurement API, and the Custom Model API, that can bring emotional intelligence to different parts of a product.

Hume's EVI API uses an empathic large language model (eLLM) to generate a conversational voice interface that can recognize and mimic human tone, word stress and speech cadence. It can be used to improve human-AI interaction by recognizing when a person is done speaking, whether they can be interrupted, and how they express emotions. And it can be tuned to match the user's response, since it's optimized to maximize happiness and satisfaction.

The Expression Measurement API can recognize subtle vocal and facial expressions in audio, video and images, and can be used to infer emotional state. It's based on 10 years of research and millions of proprietary data points, and offers high-accuracy models for gauging emotional expression.

The Custom Model API lets developers build their own insights into their products. By combining transfer learning from eLLMs and expression measurement models, the API can predict a broad range of outcomes. Examples of custom models include inferring depression in video and toxic speech in audio.

Hume's technology can be used in a variety of areas, including AI research, social networks, call center analytics, embedded devices, brand analysis, digital assistants, UX research, education and health and wellness. Developers can integrate it easily, with APIs that offer flexible and adaptable ways to add emotional intelligence to a product.

Through rigorous scientific research and large-scale training, Hume seeks to make technology more beneficial to humans. The platform produces state-of-the-art results and offers the most advanced insights into emotional expression, and the company offers support from ideation to implementation to ensure a smooth integration.

Published on August 11, 2024