I wanted to install TTS on my phone, so I downloaded the most recent APK of espeak-ng from Github. Upon opening it for the first time after installation, I get the message that the program was compiled for an earlier version of Android. Is this dangerous? I'm not sure where to find which version of Android it is compiled for. Thanks!

FWIW, I could only get RHVoice from F-droid to work. It needs to download the voice/language when you configure it, after which I disabled the network connection. It has limited voices but works well enough for me.

    Hulk Thanks, that's very helpful. I see that the same package on F-Droid has a target SDK of 26. I trust that espeak-NG itself is not malicious, but does that introduce vulnerabilities to the device?

    Also, the same page says "It is a very bad idea to enable an applications as an accessibility service through the android settings, unless you require that service to assist with a disability." Does that apply to TTS, since that is enabled through the accessibility menu?

    TranceSeven I got espeak working fine so far, although I haven't tested it aside from entering text in the application itself. IIRC, I think the radio button for "preferred engine" was already selected, but you have to select it manually to be prompted to download a voice.

      F-Droid has some issues: https://wonderfall.dev/fdroid-issues/

      Yes, Text-To-Speech is an accessibility service.
      Yes, all accessibility services introduce huge attack surfaces.

      If you must use an accessibility service, then you have no choice.

        Hulk
        Thanks for explaining!

        Yes, all accessibility services introduce huge attack surfaces.

        To the entire device or just the profile they're in?

        • Hulk replied to this.

          I'll add this: I only installed RHVoice because Organic Maps doesn't include its own TTS engine and I've been trying the app out.

          However, OSMAnd+ seems to have its own speech engine so has no need of any 3rd party TTS engine.

          Hulk Is this true with TalkBack disabled? Because I am only using TTS for Organic Maps (although I may end up using OsmAnd).

          With the screen reader on, how an attack would function seems pretty intuitive: as the warning says, "this speech synthesis engine may be able to collect all the text that will be spoken, including personal data…," etc.

          But without the screen reader on, isn't the TTS engine only getting text I've enabled an application to send to it, so I'm trusting the engine with that text as opposed to everything? I could be completely misunderstanding all of this, as I'm an absolute beginner.