Skip to content

Generalize phonemization package to support other langauges #711

@IgorSwat

Description

@IgorSwat

Problem description

As it stands for now, the Phonemis library (utilized to phonemize the input text for Kokoro model) supports 2 languages: 🇺🇸 American English and 🇬🇧 British English.

Hovewer, the utilized phonemization pipeline is flexible and should, in theory, allow to handle other languages with similar properties (syllabicability) just as well as English.

The first step to achieve that is to restructurise the code to make it easier to add new languages implementations.

What should be done

  • Design a general pipeline interface for all languages, where each language pipeline is a concrete and hermetic implementation of the interface.
  • Restructurize the current english implementation according to the design, with possibly minimal amount of API changes.

Benefits to React Native ExecuTorch

Opens the door to supporting many new languages in Kokoro’s Text-to-Speech implementation, and potentially in future TTS models.

Metadata

Metadata

Assignees

Labels

3rd party packageIssue related to 3rd party packages, but not ExecuTorch, e.g. ExpoideaNew idea to enhance the library, suggestion, etc.refactoring

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions