In the previous chapters, we talked about text-to-speech AI, voice cloning, and deepfake voice. In this article, we’ll talk about synthetic voice and where it fits in the conversation. Continue to read or jump around to learn about:
- What’s a synthetic voice?
- How do you create a custom artificial voice?
- What are the potential use cases for synthetic voices?
- How is Veritone MARVEL.ai building the future of synthetic voices?
What is a synthetic voice?
A synthetic voice is a term that fits under the umbrella of what is commonly called “deepfake” voice. It is often used interchangeably with voice cloning as well. But to put it simply, synthetic voice is computer-generated speech, also called speech synthesis, and is typically achieved with artificial intelligence (AI) and deep learning.
Two modalities are used to create synthetic voices. These are:
- Text to Speech (TTS): while we covered this extensively in our last article, text to speech is one of the ways that a synthetic voice output is generated. TTS is the process of developing human speech artificially. TTS software has been used to help read digital text for the visually impaired and in other applications such as voice bots or assistants.
- Speech to speech (STS): rather than using text, this modality uses speech to achieve the same aim, to create realistic sounding synthetic voice outputs.
TTS in the past could not generate authentic-sounding voices. But as technology has developed, that has changed. And people today can even create custom synthetic voices that they can use in various ways, which we will dive into more later.
Some of the advantages that synthetic voice has, or will, yield in terms of content creation include:
- Cost-effectiveness: voice talent and businesses alike can reduce time, minimize travel and production costs, and translate text content into a human-sounding voice quickly (even in other languages)
- Reduce studio time: before, scheduling and traveling to and from the studio ate up a lot of time and money. With synthetic voices, talent can minimize their studio time, if not eliminate it in some instances.
- Scale content: content creators, especially broadcasters, can make their repeat content, such as weather reports and sports updates, more personalized, localized, and engaging for their audiences
How do you create a custom artificial voice?
When working with AI experts, you’ll need to provide the necessary training data in audio recordings to have them produce a custom voice. This is the most significant lift on the user’s end.
In most cases, an inventory of sentences that the user will have to record them saying will be provided by the vendor. This inventory may seem daunting as it will contain thousands of sentences to ensure that the most detailed data is captured to produce an authentic-sounding voice. All in all, the recording process can take up to 2 hours, if not more, depending on the person. But once it’s done, you won’t have to worry about recording anything else.
The process outlined in the previous paragraph is only applicable if you are starting from scratch. However, if you have produced a lot of audio content and have those recordings, these can be used as a suitable substitute to save you the time recording yourself all over again. Once this data has been collected and the quality is up to par, the voice creation process can take days to weeks, depending on the project’s circumstances.
Potential use cases for synthetic voices
The use cases for synthetic media are still emerging, but we are seeing a lot of areas where people and organizations can benefit from this technology.
- Content creators: the pandemic sparked a surge in content consumption. One of the mediums that benefited from this boom was podcasting, which has grown exponentially year of year and reaching even broader, more diverse audiences. In addition, synthetic voice is already being used to help translate content in demand into different languages.
- Voice talent: many in the voice talent community are hesitant about this new technology because they fear being replaced by a robot. But this has the potential to enhance and expand their opportunities even further.
- Celebrities: well-known athletes, influencers, and celebrity voices are always sought after by the largest brands in the world. Due to tight schedules, many must decline sponsorships and advertising opportunities. But that changes with synthetic voice content, enabling them to take on more work even during their busiest time of the year.
- Advertisers: seeking voices that resonate with their target audience, synthetic voices help advertisers generate more engaging content without having to coordinate as many moving pieces such as travel and studio time.
How Veritone MARVEL.ai is building the future of synthetic voices
Leveraging our expertise in artificial intelligence and experience building solutions and applications, Veritone has made its mission to enable individual creators and organizations alike the means to create synthetic content.
With an end-to-end voice as a service application, supported by our foundation Enterprise AI platform, AI, MARVEL.ai, and our associated synthetic voice solutions help make the ethical use of synthetic voice possible.
By building rights protections for our users in MARVEL.ai, we want to lead from the front to standardize the approval and agreement process so that everyone can protect their personal synthetic voice.