evvdroid 0.2.0

This release adds every language the engine has, so it is a much larger download than 0.1.0 and now comes as one file per architecture, plus a universal one that installs anywhere.

• Eloquence now speaks ten languages rather than one: US and British English, Castilian and Mexican Spanish, French of France and of Canada, German, Italian, Polish and Japanese.
• Every language comes with the same eight voices, so you can have Reed in Polish or Glen in German.
• Lists of choices in the settings screen now scroll, so you can reach every language on a small screen.
• Pitch and rate no longer stay where they were after your screen reader interrupts.
• The Listen button in your text to speech settings now speaks in the language you picked, rather than always in English.
• The settings screen has a new Preview language row, so you can hear a voice in the language you will be reading in.

evvdroid 0.1.0

Eloquence as a text-to-speech engine for Android and Wear OS. The speech comes from openevv, a C rebuild of IBM's Embedded ViaVoice.

First release. US English, eight preset voices, all eight voice parameters, pronunciation dictionaries, and the speech rate and pitch from your system settings.

Read this before you install it

This is a hobby project. It is not from IBM, Nuance, Cerence, Google or the openevv project, and it is not a supported screen reader engine.

If you set it as your TalkBack voice and it stops talking, you have to undo that yourself: TalkBack settings, Text-to-speech settings, Preferred engine, Google. Know where that is before you turn it on.

Installing

One APK, all three ABIs (arm64-v8a, armeabi-v7a, x86_64). Android 6 and later. Sideload it, then TalkBack settings, Text-to-speech settings, Preferred engine, Eloquence (openevv). The gear beside it opens the voice settings, and so does the launcher icon.

It runs on a watch too. Tested on a Pixel Watch 4.

Notes

Speech works on the lock screen and through a locked boot, so a screen reader has a voice before the phone has been unlocked.

Settings are per voice, except speed, which is yours and stays put when you change voice. The sliders read as one control each and move by one percent from the keyboard.

No pronunciation dictionary is shipped. The settings screen will read the text ones people share.