TextToSpeech gives out a robotic voice on some Samsung devices

[OyaCanli]

Summary :

While using TextToSpeech api on some Samsung devices such as:
-Samsung Galaxy Note Fan Edition / ONEUI 1 Android 9
-Samsung Galaxy Note 9 / OneUI2 Android 10
It gives out a robotic voice.

Related api:

TextToSpeech api - attention: I'm talking about the native TextToSpeech api which uses the TextToSpeech engine of the device, not the Google Cloud TextToSpeech api

Problematical devices:

-Samsung Galaxy Note Fan Edition / ONEUI 1 Android 9
-Samsung Galaxy Note 9 / OneUI2 Android 10

Procedure

I pass Google TextToSpeech engine to the TextToSpeech constructor to define it as the preferred engine. This works as long as Google TTS is installed on the device. Otherwise it uses the default TTS engine of the device. Samsung has its own TTS engine and sets it as the default engine. Google TTS is not preinstalled on Samsung devices. There are several other TTS engines, but they are less common. I occasionnaly see PICO and IVONA among users' TTS engines

Expected result / Result on other devices

Most of the Android devices have Google TextToSpeech engine preinstalled and it gives out a nice, natural voice. Most of the Samsung users reported a fine voice as well.

Results on problematic devices

Some Samsung users reported a robotic, super fast voice.

Solutions / Workarounds :

I asked for feedback on voice quality to users who don't have Google TTS installed. I received some stats with Firebase Analytics. And approximately 6% of my total users reported a bad voice quality.

Some possible solutions:

1. Replace native TextToSpeech api with Google Cloud's TextToSpeech service or Microsoft Azure's TextToSpeech services. In that case, it won't depend on the local TTS engine on the device. But these services are paid. Alternatively, use these services only for users who reported a bad voice quality and hope not to pass free tier.
2. Use audio files instead of TTS (but since it effects only 6% of my users, I don't want to impose a bulky app to my other users)
3. Propose to users who reported a bad voice quality to install Google TTS (which is free). Of course, it is up to them to accept or not.
4. Create another version of the app with audio files and propose it to the users who reported a bad voice quality.
5. Create a dynamic module containing audio files and propose it to the users who reported a bad voice quality.

PS. In order to detect Samsung users or non-Google users, you can check it like:

textToSpeech.defaultEngine == "com.samsung.SMT" //Those who use Samsung TTS
textToSpeech.defaultEngine != "com.google.android.tts" //Those who don't use Google TTS