Real-time timbre transfer AI
This product performs timbre transfer based on real-time sound synthesis technology, which was previously considered challenging for deep learning models. It can convert any input sound into a target sound timbre such as drums, saxophone, voices etc. It can also be used to apply pre-trained effects such as reverb and distortion, or to remove them.
Feature
Converts in real-time at studio grade sound quality (48kHz)
Real-time AI voice processing technology converts input audio in real-time to any desired timbre at studio grade sound quality (48kHz)
Without additional training of the models
Timbre transfer of any input sound to a target style without additional training of the models
Feature
Feature
Encompasses a wide range of potential applications
Operates in real-time on general purpose CPUs and encompasses a wide range of potential applications, including instrumental sounds, animals, voice, effectors, effect elimination, etc.
Use Case
Creative performance
Creative performances that would have never been possible before, such as transforming a performing musical instrument into a completely different sound, or transforming one’s voice into that of a celebrity is now possible.
Acoustic environment modification
Automatically correct acoustic environment such as noise and echoes that have been difficult to remove using conventional methods.
Implementation
We offer Neutone Morpho, an audio plug-in that enables real-time timbre conversion and acoustic processing. For more information, please visit the project website
Samples
Example of real-time conversion of drum sounds into human voices
Example of real-time conversion of piano sounds into hymn-style/sutra-style
Technology
This technology consists of multiple components, the plugin for real-time audio processing and its specialised ability to deploy AI models for processing using neural networks. The plugin is compatible with the VST3 and AU formats and can be used with general-purpose DAW (Digital Audio Workstation) interfaces and other audio software environments. The AI model for timbre transfer applies a learning architecture that combines Variational Autoencoders with adversarial learning. The encoder layer expands the input sound into the latent space, “interprets” it, and then controls the decoder layer to reconstruct the target sound.
Tech Spec
Price System
License term: Monthly Developer's license: Yes
Input/Output
Input: Audio Output: Audio
Operating Environment
Provided as a program that is designed to run in a local environment
Processing Speed
Real-time
Other products