Custom Models

Combined with words, expressions provide a wealth of information about our state of mind in any given context like customer satisfaction or frustration, patient health and well-being, student comprehension and confusion, and so much more.

Hume’s Custom Models API unlocks these insights at the click of a button, integrating patterns of facial expression, vocal expression, and language into a single custom model to predict whatever outcome you specify. This works by taking advantage not only of our state-of-the-art expression AI models, but also specialized language-expression embeddings that we have trained on conversational data.

The algorithm that drives our Custom Models API is pretrained on huge volumes of data. That means it already recognizes most patterns of expression and language that people form. All you have to do is add your labels.

You can access our Custom Models API through our no code platform detailed in the next section or through our API. Once you create your initial labeled dataset, your labels will be used to train a custom model that you own and only your account can access. You’ll be able to run the model on any new file through our Playground and Custom Models API. You’ll also get statistics on the accuracy of your custom model.

---