Voice artificial intelligence is an emerging technology, and even industry insiders have failed to agree on a coherent description. It is used by technology bloggers to describe any interaction of artificial intelligence and automated voice recognition (how computers interpret spoken language) and/or text to speech production (how computers speak).
Some writers refer to smart speaker personas like Alexa as “voice AI.” Others use the phrase to describe synthetic voice production based on machine learning. This gap in usage implies two definitions. With that in mind, here’s a suggested entry for a future dictionary.
How Businesses Are Using Voice AI Assistants?
Most brands will employ voice artificial intelligence in the second sense of the preceding statement. That is, they will create a literal brand voice using a deep neural network from a supplier such as ReadSpeaker (which we will describe shortly).
However, as the Voicebot list shows, prominent brands may end up developing their own virtual assistants and/or smart products to host such identities. Below are some instances of branded voice AI assistants:
- In 2020, Bank of America introduced Erica, an AI-powered virtual financial assistant. As of March 2021, Erica had performed over 45 million “client requests” from 8 million users, which ranged from reading out a customer’s routing number to tracking down individual transactions to notifying that a recurring charge amount had changed—all via a voice user interface.
- The previous year, Capital One released Eno, an AI helper. Eno was one of the first branded voice bots outside of the main smart speaker avatars. The virtual assistant is available on Capital One’s mobile app and website.
- Drivers in new Mercedes cars can activate the native AI virtual assistant by saying, “Hey Mercedes.” This system recognizes various commands using natural language comprehension, which is a type of artificial intelligence.
Artificial Intelligence and Synthetic Voice Generation
Artificial intelligence enables the development of lifelike synthetic voices, such as AI voice clones that closely resemble the sound of a certain speaker. Engineers employ deep neural networks (DNN) to develop an AI voice clone, which is a complicated type of computer architecture that mimics synaptic connections in the human brain. These systems identify patterns in data sets. That means you can train them and they will “learn.” The process of training a model on a DNN is known as deep learning.
To voice clone, technicians feed audio recordings of the original speaker into deep voice software, a form of neural network. The DNN recognizes the minute patterns in that voice—tone, pronunciation, pace, stress, rhythm—and produces a model that can emulate those intricacies while executing completely.
1. Artificial intelligence voice clones for celebrity spokespersons.
In the early to mid-2000s, actor James Earl Jones was “the voice of Verizon.” He was featured in the company’s advertising. He held live branding activities. Back then, there were few voice-based touchpoints between brands and their customers, so Jones’ recording schedule was doable.
If Verizon and Jones had the same relationship today, the corporation would go bankrupt paying the actor to record scripts for all of the new voice-based channels, including commercials, smart speaker apps, interactive voice response (IVR) systems, and so on. A licensed James Earl Jones voice clone would allow Verizon to keep its branding consistent across all voice channels without incurring the cost or scheduling problems of numerous recording sessions.
2. Consistent Brand Mascot Voices.
Voice cloning is not limited to real-life celebrities. Characters like Ronald McDonald, Mickey Mouse, and Chester Cheetah help to establish a consistent brand experience across all audio channels. Voice cloning enables a character’s voice to remain consistent throughout generations, eliminating the small changes that occur with swapping voice performers.
Conclusion
When brands create custom voices, they can use this unique identifier on a growing number of voice-first devices and media, including voice AI assistants, IVR systems, in-car infotainment, interactive in-store displays, e-learning materials, TV, radio, online ads, instruction videos, accessibility tools, conversational robots, and more. This results in a consistent experience that follows the client throughout the day, increasing awareness, trust, and loyalty—all without the expense of continuous talent contracts.
This is the type of speech artificial intelligence that will propel brand innovation forward. Voicebot.ai’s 2024 list of top voice-branding efforts does not include many AI-generated voices. That year, creating a smart speaker app was sufficient.
