Closed ihateprogramming88 closed 1 week ago
Please use a good neural network based engine so it wouldn't be a robotic and annoying voice reading the text. There are numerous open source read aloud engines out there that you can use.
Does anyone have any recommendations for neural network engines that can be used?
A simple google search brought me to this:
https://github.com/espeak-ng/espeak-ng
They say it's lightweight and low footprint. It even has an android implementation for testing which I tested and it was OK. Its development doesn't seem to be dead as the last commit was 8 days ago.
But that's just my search results, and I am not a developer myself. Please probe this project and search for other TTS models to find the one most suitable.
This one also seems interesting: https://github.com/coqui-ai/TTS
Android has its own API called Google Speech Recognition & Synthesis. Third party apps can use the built in API to get accessibility features. But it is not as good as other neural models. And also, it relies on the provided API from android, which means if you're using a custom ROM android that ripped this API out, the ReadYou app cannot produce TTS on its own.
This one also has good results:
https://github.com/rendchevi/nix-tts
But its development seems dead.
Mozilla has its own TTS engine too:
https://github.com/mozilla/TTS
It's also good but I don't know if it's lightweight or suitable for this job. Pocket text to speech feature is good, but I couldn't find any evidence that Pocket is using Mozilla's TTS model or not.
This one also didn't see the charm of development in 3 years. I don't know how that would effect you or the implementation. But you can always train the model yourself I guess.
Reeder is another FOSS RSS reader that does TTS read aloud pretty well, but I'm not sure what library is being used, and there's only one voice: https://github.com/spacecowboy/Feeder
Another engine to look into is Piper TTS: https://github.com/rhasspy/piper
My only other suggestion, from a user experience perspective, would be to highlight the paragraph being read, and having the ability to click on a paragraph to go back to it. This way users can replay or skip to the sections of their choice.
Here are some examples from other reading apps:
Great, thanks for the input, I'll do some research and decide what is best suited. Hopefully will get started soon.
Temporary solution: Utilize the ability to read aloud in webview.
Awesome, nice stuff, I have downloaded the latest build, how do I use read aloud. I can see bionic reading is working nicely. Just can't figure out how to use read aloud, WebView is enabled. @Ashinch
Awesome, nice stuff, I have downloaded the latest build, how do I use read aloud. I can see bionic reading is working nicely. Just can't figure out how to use read aloud, WebView is enabled. @Ashinch
In Webview, needs select text and expand text tool to tap "Read aloud" button.
I‘m looking for a more convenient way to trigger.
I can't seem to get this to work on GrapheneOS, do you need any extra applications installed, I have google speech synthesis installed but when I delete the text in WebView no option pops up with read aloud at the moment on latest nightly build.
Not particularly a feature request, I wanted to work on the read aloud feature. Is there any particular way that you want this implemented?