Compute-efficient Real-time Voice Cloning
By
A voice cloning machine learning (ML) model receives a speech and text input and creates a new speech output reading the text input in the voice of the speaker. Our project aims to both speed up processing and reduce the computational resources necessary to run a voice cloning ML model, which can then be uploaded to a low-end system. The project uses a pre-existing machine learning toolkit repository to speed up the productivity of machine learning engineers. By implementing a modified ML model with updated sub-models into the existing model, we gain access to an improved training and evaluation environment that is more accessible to a broader audience. Once the updated model is complete, it can be implemented on a low-end system for user interaction. The user peripherals consist of a miniature button keyboard with an attachable display and a microphone for user inputs, and a volume adjusted amplifier for the cloned voice output.
Artifacts
Name | Description | |
---|---|---|
Executive Summary | Project summary and motivation for design. Includes project timeline. | Download |
Project Document | A document containing documentation for the entire Compute-efficient Real-time Voice Cloning project design process. | Download |
Project Summary Video | A summary video for the Compute-efficient Real-time Voice Cloning project. | Link |
GitHub Repository | GitHub Repository for our code. | Link |