This repository contains the information about the data packages, that is the language and voice packages.

The information about each language or voice is stored in one or more separate json files (see the src/ directory).

The build script creates a single json file merging all the information. This file itself is not tracked by git, but when one of the source files is updated and the update is approved, the newly generated single json will be uploaded to our server. This file will be downloaded by the app for Android and hopefully other clients in the future, so no applications need to be recompiled and updated on the users' devices to give them access to the up-to-date versions of the language and voice packages, unless those packages require features only available in a later version of the RHVoice core.

Any updates first appear in this branch. No users will have access to them at this point. This is for testing by developers.

When an update has been tested by the developers, it will be cherry-picked into the alpha branch. The json file built from this branch is used by the Alpha version of the RHVoice app, so the updates to this branch will be available to the testers using that version of the app.

When the update is considered ready for general use, it will be cherry-picked into one or more of the stable branches. The only such branch so far is 1.8. But future updates of the RHVoice core may make it necessary to introduce new stable branches. Contributors will need to document the usage of any core features introduced after RHVoice v1.8.

The production version of the app will contain the latest core and use the data from the most recent stable branch.

If you want to publish a new voice or an update to the existing language or voice, follow these steps.

Packages are simply zip archives of the data. Everything in your data/languages/<language> or data/voices/<voice> directory goes into the archive. If you compile under Linux or Windows, you will find a ready-made archive in build/linux/packages/zip or build/windows/packages/zip.

You need to provide or arrange the hosting for your packages.

The server hosting your packages must support https.

If you are contributing a new voice or have updated an existing voice such that it has changed the way it sounds, synthesize the language-specific demo text and upload the clip. The audio file must be in either Ogg Vorbis or MP3 format.

If you are contributing a language, please help us extend this table with a demo message for your language. Just translate the English message.

See the existing files in the src directory for examples.

Some notes on the expected formats of values:

• Two-letter language codes are ISO 639-1 codes.
• Three-letter language codes are ISO 639-3 codes.
• Country codes are ISO 3166 codes.
• MD5 hashes are encoded as Base64. For example, you could use the following command to generate the hash, if you have openssl and base64 programs installed on your Linux system:

  openssl md5 -binary <YourPackageFile> | base64
  

  Also, you can use the md5_hash Python script from the packages directory. To use the md5_hash Python script, you need to specify a directory containing language and voice packages, and to specify a file name where MD5 hashes will be stored, (EG)

  md5_hash </home/packages> </home/md5_hashes.txt>
  

  From Windows:

  python md5_hash <D:\packages> <D:\md5_hashes.txt>
  

Let's imagine your two-letter language code is xy. And, if you have created or updated a voice, you've saved the information about it in src/languages/xy/voices/myvoice.json.

Now let's run the check.

If you are publishing a language, run: ./check xy

If you are publishing a voice, run: ./check xy myvoice

The script will do some basic checks, including downloading your package and verifying that the hash matches.

Before your contribution is made available to the Alpha testers, you must test it on an Android device and confirm that it works as expected and causes no major issues, such as an RHVoice crash.

1. Build the package directory: ./build
2. Upload the packages-dev.json, created by the previous command, somewhere.
3. Setup the build environment to build the RHVoice app.
4. Create or edit ~/.gradle/gradle.properties, configuring the url to download the package directory:

   RHVoice.devPkgDirUrl=<url>
   

5. Build the development flavour of the app: ./gradlew assembleDevDebug
6. Install and test the app you have built, on an Android device with your voice.

Make a pull request to the main branch. When it's merged, the maintainer will also upload the development package directory. Make a final test of the development app build with the url pointing to that directory. The build process is the same, except the RHVoice.devPkgDirUrl property must not be set.

When you are sure your update is ready for testing by users, leave a comment confirming it in the pull request.

When the update is available in the Alpha version of the app, there will be a comment about it in the same pull request.

The testing is done as a closed test on Google Play, so you need to find testers for your voice and provide us with their gmail addresses in an email to [email protected].

When we register your testers, they will have to use the special link we will send to you to confirm to Google that they really want to become testers.

Gather feedback from the testers. If necessary and possible, fix the issues and publish the fixes.

When your update is ready for use by the general public, open an issue in this repository requesting the publication. Your package will be made available to the stable version of the app.