Tdpsola: an implementation of the tdpsola algorithm

TD-PSOLA (Time-Domain Pitch-Synchronous Overlap and Add) is an algorithm that allows to modify the pitch and the speed of a sound independently of each other.

The algorithm

The idea behind the algorithm on itself is quite simple, but hard to explain. I have found this documentation the clearest.

Advantages of TD-PSOLA:

• Preserves the formants, i.e. if you increase the pitch of a human voice, it doesn't sound like the person has been breathing from a helium balloon
• Performant algorithm for modest pitch/speed changes

Disadvantages of TD-PSOLA:

• Not robust against noise, only suitable for voiced sounds
• Requires a-priori knowledge of the fundamental frequency of the source sound

The implementation

This is an implementation in the Rust programming language. Advantages of this implementation:

• 0 samples delay
• separate analysis and synthesis, allowing you to re-use the analysis part
• speed and pitch can be manipulated independently at any time
• many tests
• implemented in a memory-safe programming language, without "unsafe" features

Disadvantages of this implementation:

• The algorithm is extra complicated in order to guarantee 0 samples delay.

Examples

A simple example can be found in the documentation or in the src/lib.rs file. A more worked-out example can be found in the examples folder in the source code.

Planned features

• suitable for real-time analysis and synthesis
• block-processing for extra speed (can currently only be used sample per sample)

Contributing

We welcome contributions, both in the form of issues and in the form of pull requests. Before opening a pull request, please open an issue first so that you know whether a subsequent pull request would likely be approved.

If you don’t have a Codeberg account, alternatively, you can contribute via e-mail (an email address is in the Cargo.toml file). Just creating a Codeberg account is probably the easiest.

Unless explicitly stated otherwise, you agree that your contributions are licensed as described below.

License

This is free software: you can redistribute it and/or modify it under the terms of the GNU Affero General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This software is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Affero General Public License for more details.

The full license text can be found in the file LICENSE-AGPL-3.0.md that should be distributed with this source code. If it is not there, see https://www.gnu.org/licenses/.