Generative AI

Build a Digital Human Interface for AI Apps with an NVIDIA NIM Agent Blueprint

Providing customers with quality service remains a top priority for businesses across industries, from answering questions and troubleshooting issues to facilitating online orders. As businesses scale operations and expand offerings globally to compete, the demand for seamless customer service grows exponentially. 

Searching knowledge base articles or navigating complex phone trees can be a useful resource, but the density of information customers must often parse through can contribute to poor retention.

With the convergence of technologies such as generative AI, conversational AI, and visual AI, application development teams are seeking ways to increase engagement with self-serve customer applications to boost overall satisfaction and retention. 

Instead of a text-based chatbot, you can now add digital human interfaces to personalize chatbot applications and use retrieval-augmented generation (RAG) to offer a smooth, human-like interaction. This post explains how this can be achieved using the digital human NVIDIA NIM Agent Blueprint, which combines NVIDIA NIM microservices with reference code and documentation.

NVIDIA NIM Agent Blueprints are reference workflows that enterprises can use to build and operationalize custom AI applications. NVIDIA NIM is a set of easy-to-use microservices designed to accelerate deployment of generative AI across cloud, data center, and workstations. 

Designing a human-like chatbot for customer service

When designing a customer service tool, it’s important to consider your end users and their desired method of interfacing with your application.

While text-based apps offer several advantages, the need for multilingual language support and accurate information retrieval fuel the demand for digital human interfaces.

For example, James is an interactive digital human who is knowledgeable about NVIDIA and NVIDIA products. Enterprises looking to provide scalable, engaging, customer service support can leverage a similar avatar to fulfill these needs with higher accuracy using RAG for information retrieval—all while operating with low latency.

To build better user interfaces that address diverse user needs, you can use the digital human for customer service NVIDIA NIM Agent blueprint, which uses the following: 

  • NVIDIA Riva ASR NIM: State-of-the-art Parakeet automatic speech recognition (ASR) model, capable of transcribing spoken English with exceptional accuracy.
  • NVIDIA Riva TTS NIM: FastPitch is a mel-spectrogram generator, designed to be used as the first part of a neural text-to-speech (TTS) system in conjunction with a neural vocoder. This model can output a female or a male voice for US English.
  • NVIDIA Audio2Face NIM: Animates 3D character’s facial characteristics to match any audio track. 
  • Llama 3 8B NIM: Advanced state-of-the-art large language model (LLM) with language understanding, superior reasoning, and text generation.

For more information, visit NVIDIA ACE.

Video 1. See how digital humans are transforming industries with NVIDIA ACE

A digital human NIM Agent Blueprint for customer service apps 

The digital human for customer service NIM Agent Blueprint is powered by NVIDIA Tokkio, which is a reference workflow of the ACE ecosystem. It brings enterprise applications to life with a 3D animated digital human interface. With an approachable, human-like interface, customer service applications can provide better user experiences with faster resolutions than traditional customer service options.

This workflow is designed to integrate within your existing generative AI applications built using RAG (Figure 1). 

Architecture diagram showing digital human workflow with audio, video, and agent capabilities.
Figure 1. The digital human for customer service NVIDIA NIM Agent Blueprint diagram

There are two main steps to evolve your applications to include a full digital human interface, whether they’re running in your data center, in the cloud, or at the edge.

Step 1: Initiating user interaction 

Walking through the workflow shown in Figure 1, audio from the user comes in through the web front end and is processed by an audio/video engine and passed on to the NVIDIA ACE agent.

Step 2: Powering life-like digital humans

The agent uses the audio pipeline to convert audio to text and text to audio as it interacts with the RAG-powered chatbot or copilot using its API. The audio is sent to the 3D animation pipeline to animate the avatar and render its features to be lifelike. 

As users engage the digital human, they can provide feedback on the response (thumbs up or down, for example), which is then fed into the backend RAG application for improvement.

The digital human for customer service NVIDIA NIM Agent Blueprint includes all the components you need to get started with your own digital human project:

  • Digital human, named Ben, which can be customized for your applications
  • Sample applications 
  • Customization documentation
  • Reference code
  • Helm chart
  • Integration guidelines
  • Deployment instructions
  • Evaluation metrics

The comprehensive package is designed to help you quickly create and deploy AI-powered digital humans for customer service applications, while enabling customization to meet your specific business needs. 

By leveraging this NIM Agent Blueprint and its components, you can:

  • Increase engagement and satisfaction for user-facing applications by enabling natural conversations with enterprise applications and data.
  • Use a suite of easy-to-use and performance-optimized NVIDIA NIM microservices, for avatar animation, speech AI, and generative AI.
  • Create lifelike digital humans that are rendered with subsurface scattering for accurate skin and hair through the Omniverse RTX microservice, animated with the Audio2Face NIM, and has a responsive speech interface with the NVIDIA Riva Parakeet NIM and ElevenLabs integrations.
  • Runs anywhere, in an on-premises data center, in the cloud, or at the edge.

Get started

Experience the digital human for customer service interactive demo. Navigate to the Blueprint Card to see customization, compatibility, and deployment information.

To see the digital human for customer service NIM Agent Blueprint documentation, sample code, and Helm chart, visit the Generative AI Examples GitHub repository.

To start building applications using NVIDIA NIM microservices, visit the NVIDIA API Catalog. You’ll be prompted to join the NVIDIA Developer Program for free access during development. Or request a 90-day trial of NVIDIA AI Enterprise, including access to NIM with NVIDIA support. For more information, see the NVIDIA NIM FAQ.

Discuss (0)

Tags