EA Working On Aging Characters’ Voices With Game Progression

Rundown:

• Electronic Arts has published a recent patent aimed at making character voices in video games sound realistically appropriate for their ages.
• The patent outlines a system that employs machine learning to alter voice audio in video games as characters age in the storyline.
• The process involves inputting the initial audio signal and an “age embedding” into a machine-learned model.
• The age embedding is based on the age classification of speech audio samples, helping the model generate age-appropriate voice changes.
• The system employs various machine learning models to transform the original voice into one that matches the chosen age and gender of the character, ensuring realistic and coherent character development.
• The system also aims to continually improve by comparing the altered voice with actual recordings of people of the chosen age, adjusting itself to achieve better results in the future.

Earlier today, we came across a recently published patent from Electronic Arts titled, “Voice aging using machine learning,” filed in August 2021 under the name of Electronic Arts Inc. The patent, published earlier this week, describes a system for making voice audio in video games sound like it belongs to characters of different ages.

As people get older, their voices change. Electronic Arts wants to make sure that the voices of characters in video games also change realistically as they age in the video game’s storyline. Imagine you’re playing a video game where your character grows older as the video game progresses. It would be strange if the character’s voice didn’t change along with their age. This technology solves that problem by using machine learning.

“This specification describes systems and methods for aging voice audio, in particular voice audio in computer games. According to one aspect of this specification, there is described a method for aging speech audio data,” reads the abstract for the patent.

“The method comprises: inputting an initial audio signal and an age embedding into a machine-learned age convertor model, wherein: the initial audio signal comprises speech audio; and the age embedding is based on an age classification of a plurality of speech audio samples of subjects in a target age category; processing, by the machine-learned age convertor model, the initial audio signal and the age embedding to generate an age-altered audio signal, wherein the age-altered audio signal corresponds to a version of the initial audio signal in the target age category; and outputting, from the machine-learned age convertor model, the age-altered audio signal.”

First, the system takes the original voice audio of a character and some information about the character’s age. It then uses a special model to figure out how old the character should sound. This model listens to lots of different voice samples from people of different ages and learns to tell their ages based on their voices.

Using the original voice and the age information, the system uses another machine learning model to change the voice so that it sounds like it belongs to someone of the desired age. It gives you a new voice that sounds like the character at the specific age you want.

This patent also talks about how the age information is generated, how the models learn to make the voice changes, and how the models can be trained to get better at it. It also mentions that this system can work with different types of machine learning models and can take into account other factors like gender to make the voice sound even more realistic.

According to the patent’s claims, players start with an original voice recording and some age information. The age information is like a special code that comes from a model that knows how to guess someone’s age based on their voice.

Using this information, the system changes the original voice recording to sound like it’s from the chosen age group. The new voice recording is then ready to be used. This model has different parts, and the age information comes from one of those parts.

The age information is kind of like a bunch of numbers that represent the age. These numbers are made by looking at many voice samples from people of a certain age group. Players can start with written words and turn them into voice recordings using the system.

This is helpful when the text is something a character in a video game would say. These voice recordings are used as the starting point for this age-changing process. There are different types of machine learning models that can be used to make voices sound different ages.

Players can also give the system information about the character’s gender. Then the changed voice recording will match both the chosen age and gender. The system also compares the new recording with actual voice recordings from people of that age. If they’re not similar, it adjusts itself to do better next time.

Electronic Arts aims to introduce a transformative era for video game narratives by seamlessly integrating evolving character voices into the gameplay. The system’s capacity to dynamically adjust voices based on age and gender significantly elevates the immersive quality of virtual worlds, ensuring that players remain captivated by the unfolding storylines.

As the video game industry continues to push boundaries, this patent demonstrates the convergence of cutting-edge machine learning and gaming innovation. Players can look forward to a future where characters’ voices authentically reflect their virtual journeys, bringing an unprecedented level of engagement to their gaming experiences.

Importantly, it should be acknowledged that the publication of this patent does not assure its development or integration. Only time will reveal how Electronic Arts plans to include this system in its current and future video game offerings.

What do you think about this? Do tell us your opinions in the comments below!

Similar Reads: Sony May Allow Players To Merge Faces With In-Game Characters