Voice Aging Using Machine Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In video games, the lack of age progression in character voices can break immersion, and storing multiple versions of speech for characters at different ages is memory inefficient.
Innovation Solution
A method using a machine-learned age convertor model that processes initial audio signals and age embeddings to generate age-altered audio signals, employing models like autoregressive sequence-to-sequence or LSTM models, to create realistic voice aging, while also considering gender differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple versions of speech are stored for characters at different ages, then voice aging realism is improved, but memory usage increases
Solution Approach 1:
The patent creates a single base speech recording and uses machine learning models to generate multiple age variations as copies. Instead of storing multiple original recordings, the system generates synthesized copies of the same speech at different ages, dramatically reducing memory requirements while maintaining voice aging realism.
Solution Approach 2:
The system changes the age parameter of a single speech recording to generate multiple age versions. By adjusting age-related acoustic parameters through machine learning, the same speech content can be rendered at different ages without storing multiple complete audio files.
2Quantity of substance
If a single speech recording is used for characters of all ages, then memory usage is reduced, but voice aging realism deteriorates
Solution Approach 1:
The patent replaces the mechanical approach of storing multiple physical audio files with a computational approach using machine learning models. The system substitutes actual multiple recordings with synthesized generations from a single recording, achieving both memory efficiency and voice aging realism through algorithmic transformation.
Solution Approach 2:
The machine learning model acts as an intermediary between the single base speech recording and the multiple age variations. This intermediary processes the original recording and generates age-appropriate transformations, enabling realistic voice aging without requiring multiple source recordings.
3Reliability
If manual voice recording for each age is performed, then voice aging quality is improved, but production time and cost increase
Solution Approach 1:
The system enables self-service voice aging where the machine learning model automatically generates age variations without requiring manual intervention for each age group. The model processes the base recording and autonomously creates multiple age versions, eliminating the need for repeated studio sessions and manual recording for each age category.
Solution Approach 2:
The patent performs preliminary action by creating a single high-quality base speech recording that serves as the foundation for all future age variations. This preliminary recording captures the essential voice characteristics, and subsequent age transformations are generated computationally, avoiding the need for repeated preliminary recording sessions for each age group.
Data Source
AI summary
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. 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.