Voice User

Cobalt Interact TM

Human-like capable Voice Interface

Cobalt Interact allows defining sophisticated conversational voice flows to create a powerful interactive natural-language system. Cobalt Interact provides a seamless interface between the ASR speech recognizer, Cobalt Understand (NLU) engine, and Cobalt VoiceGen, our text-to-speech (TTS) engine. The result is a highly configurable spoken-language understanding user interface that can easily integrate with your back-end service solutions.

Use Cases / Examples

Cobalt Interact can be used to create spoken interfaces
in a wide variety of applications.
Examples include:

Control & query of complex apps & devices

“Copier, make 20 double sided copies”
“Toby, what’s the engine temperature?”

An interactive, talking user manual or knowledge management system

“OK LiveBook, How do I change the oil?”
“How do I set the recording time?”

An automated retail concierge.

“Hey Shoppy, where do I find pickles?”
“Does this shirt come in a V-neck?

A help system for customer support

“How do I add a driver to my policy?”
“I need to change my address."

Natural language database queries

“OK Tutor, what’s the capital of Bolivia?”
“What’s the exchange rate of the Euro?”

Always Listening, Yet Private

Cobalt Interact can be configured with your own “wake word” to trigger the system, and since it can run entirely locally, users don’t have to worry about their speech ending up in the hands of a cloud-based speech company.

Always Your Brand’s Voice

Cobalt Interact can speak with a custom voice that helps identify and distinguish your corporate brand.


A fortune 250 company that operates multiple insurance businesses wanted to deliver a better user experience to its clients while reducing the volume of incoming calls into their contact center with better self-services offerings


The company used Cobalt’s mobile voice interface through which the company’s clients could file & follow up claims from the app or portal without having to place a call into the call center.


The solution reduced call volume dramatically while boosting customer experience and engagements 


NLP (VUI) Voice User Interfaces: Insurance Claims

Cobalt VoiceGen TM

Synthetic Voices for devices, sites or apps

Cobalt VoiceGen, our Text to Speech (TTS) engine, uses deep learning-trained parametric synthesis models to produce speech that sounds natural, highly intelligible and customizable.

Earlier Text-to-Speech (TTS) technology was concatenative, piecing together syllables and sound fragments to create a new synthetic voice, which sounded stilted and robotic . Cobalt’s VoiceGen uses Deep Neural Networks, creating a much more natural-sounding generated voice that can change speed, emphasis, and even emotions.

Understand TM

Natural Language Understanding

Cobalt Understand (NLU), takes in a string of text and extracts two pieces of information: intent and entities. It uses a flexible model pipeline to allow chaining together exact-match rules, keyword-based rules, and deep-learning models for optimal accuracy. 

Using machine learning, Cobalt Understand extracts things like key phrases, topics, sentiments and entities to uncover valuable insights from documents like emails, insurance claims and customer support tickets.

Cobalt Interact Model Updater TM

Update Your ConversationalAI models

Cobalt’s Interact system allows you to craft sophisticated dialogue flows called Stories to map what users say to what they intend to accomplish.  The Cobalt Interact Model Updater transforms these human-readable Story files into customized automatic speech recognition (ASR) and natural language understanding (NLU) models that Cobalt’s engines use at runtime. Unlike many systems that use a generic ASR model, Cobalt creates a model specifically adapted to the speech patterns you expect from your users to maximize accuracy.