news?

Any news? tried WebSpeech (works) with WebView(html)+expo-av(mic) but something is missing, it doesn’t prompt for access to the device’s microphone (SDK 41). Ideas?

Here’s a working WebSpeech API example on any mobile: https://www.google.com/intl/en/chrome/demos/speech.html

This seems like a more popular/robust choice: https://github.com/react-native-voice/voice. rclai what do you think?

For online speech recognition, I don't think expo should do anything...just use a cloud service or run it in the cloud. Really I think the benefit of going mobile is being offline (for low-latency or resiliency).

One way of doing this, while also enabling much broader functionality is tensorflow lite: https://expo.canny.io/feature-requests/p/tensorflow-or-keras-support.

The developer would have to run their own model. Offline speech recognition would require a very large model, hundreds of MB after tflite conversation and quantizing:

For simple commands, the model could be very small...mere KB even:

Android

iOS

There might be a very experimental way to do this today with a new project I spotted: https://github.com/tensorflow/tfjs/tree/master/tfjs-react-native. I'm curious if loading the model over the bridge is slow or if there are other tradeoffs vs tflite with native Android and iOS.

When I get time, I will try and run the speech command js example with tensorlfowjs with reactnative tensorflowjs. https://github.com/tensorflow/tfjs-models/tree/master/speech-commands

Great library. Guess I’ll need to eject my app.

This module is the only native module holding me back from using Expo on my app

Will this happen any time soon? Trying to decide whether to eject or use Google Cloud SST as a stop-gap solution while we wait.

Agree

Can't wait to see this feature in Expo.