world-vocoder

Main repo: https://github.com/mmorise/World

"WORLD is free software for high-quality speech analysis, manipulation and synthesis. It can estimate Fundamental frequency (F0) ([fundamental-frequency-estimation]), aperiodicity and spectral envelope and also generate the speech like input speech with only estimated parameters." - from the Github repo

It is a [vocoder] that is designed for integration into [statistical-parametric-speech-synthesis] systems.

License: modified BSD

It splits input speech into three components:

• [fundamental-frequency-estimation] (F0) based on DIO (Distributed Inline-filter Operation) or Harvest
• [spectral-envelope] estimation using [cheaptrick]
• [aperiodicity]

As I understand it, the [spectral-envelope] is the actual vocal quality that we're trying to capture + model. It consists of the different harmonics that make up the intonations of each of the pieces of speech.

wasm port

There is a JavaScript API version available via Emscripten/WebAssembly: World.JS

Example code with psuedo-notebook style: https://www.gloomyghost.com/World.JS/#/?id=synthesis_js <- this is quite good and could be a great starting point for experimentation with this

links

• hts-engine-world: A project combining the World vocoder with [hts] to implement a [statistical-parametric-speech-synthesis] engine: https://github.com/mipuc/hts-engine-world
• Another common [vocoder] used in [[parametric-speech-synthesis]], STRAIGHT: https://www.jstage.jst.go.jp/article/ast/27/6/27_6_349/_article