Can I access microphone during voice recognition? about voice

[wenkesj]

Android

To save the voice, you would need to capture the raw audio signal, that would require a very simple pull request actually!

According to a quick stack overflow search:
https://stackoverflow.com/a/16340756
"You can use the "buffer received" callback..."

This can be updated for the Android API in the following steps:

1. Visit this method in the source code
2. Convert the byte[] buffer passed from onBufferReceived(byte[] buffer) to a readable React Native buffer as a local variable
3. Add the buffer to the WritableMap object event. This would be similar to the error object event.putBoolean("error", false);, except you would add event.put<Insert Type Here>("buffer", buffer);.
4. Test it with the example app by reflecting your use case using the Voice.onSpeechRecognized event listener. Every callback concatenate the received event.bytes with a global array variable and then finally writing the contents to a file (This might require another library to write to a file similar to node fs).
5. You can possibly interface with other libraries such as react-native-audio-toolkit to read the file after you saved it.

Good Luck! I would love to see this implemented :)

[ohtangza]

For iOS, it seems that there is no port we can access the low buffer of PCM. I'll give it a try when I have some time.

[wenkesj]

@ohtangza I recognized your fork is ahead of the main repo, I cherry-picked some of your commits and am planning on merging them so you can stay up-to date as well as everyone else.

I am keeping some of the extra changes you made with tracking break-points in sentences, I think that adds some value to most apps with no performance disadvantages.

Another note is that I added a configuration option for Android clients for the RecognizerIntent on start.

[wenkesj]

@ohtangza is that okay? I will credit you as well as give you full access repository.

[ohtangza]

@wenkesj Oh, I think it's okay for you to cherry-pick our changes. Although we are using it for our app now, I'm not sure this satisfies the level of public use for distribution.

I think you can add a remote repo from your side and apply conditional cherry pick commits if you want. If I am not right, please let me know. Ah, one more thing, regarding getting buffer from Recognizer, we tried but it seems that the recent version of Android does not guarantee the callback to be fired, so we stopped there. Please let us updated if you make a another process on this issue.

@ghsdh3409 Put you in the loop because you also work on our forked repo.