search

pilot51 / voicenotify

Android app that speaks notifications
https://voicenotify.app
Apache License 2.0
166 stars 57 forks source link

Choose different TTS engine #47

Open oldherl opened 9 years ago

oldherl commented 9 years ago

Please add an option for specify a different TTS engine to use other than the default one. It can be useful when the system UI is set to one language and notifications (like messages) are often in another language. In this case, say the notification in another language (by another engine) can make the voice more reasonable.

paolo-caroni commented 1 month ago

Nice, but i think that it will be more profitable to simply pre analyze the text determining the language, for example with fasttext-language-identification and pass the text with the language flag to the engine. This will help also in the case of bilingual or multilungual people, receiving message in more than one language. Also it's non needed change engine, only language.

oldherl commented 1 month ago

Nice, but i think that it will be more profitable to simply pre analyze the text determining the language, for example with fasttext-language-identification and pass the text with the language flag to the engine. This will help also in the case of bilingual or multilungual people, receiving message in more than one language.

Maybe not. In my case, often the message is written in multiple languages. The most common case is a chat message where the name of the speaker is in a language different than the content, and the identifier would not know what to do.

Also it's non needed change engine, only language.

Wrong. Some engines are better at one language than another, so it's desirable to use different engines for different languages.

paolo-caroni commented 1 month ago

Maybe not. In my case, often the message is written in multiple languages. The most common case is a chat message where the name of the speaker is in a language different than the content, and the identifier would not know what to do.

In this case the identifier should send the name in one language or maybe read the name with the wrong accent but the text with the correct language. This depend how it is implemented, a better analisis probably will discover also a english word in a spanish/italian/french sentence. But this will need much cpu power calculation since the cited languages have latin derivation and are writed with the same "symbols". Found an isolate an english word in a cyrillic, arab or on ideograms is surely simple and need lower cpu power.