New release for Transcribable Android & Wear OS app; introduces local audio recording to the Wear OS app for offline Speech-to-text processing using the local Whisper services from the Transcribable Android application.
Table of Contents
Android OS app (369011304A)
- Library activity revised and split in to two tabs:
- Files tabs shows files from the default/custom storage location.
- Recordings tab shows audio files sent from the Wear OS device.
- These audio files can be played via a compatible player installed on the device.
- Provides the option to re-trigger transcription against an audio file. Useful when testing different Whisper models.
- Improvements to Wear OS device connectivity checks.
Changes included from open testing release (369011300A)
- Addresses a potential crash when handling text files over 1MB.
- Provides the ability to select an audio file from storage for Transcription (similar to Share to) and append results to file.
- Library Fragment now includes additional file context options:
- Rename, Delete, Duplicate and Share.
- Now uses a custom Onnx runtime build:
- Silero VAD format now uses ort format (app will prompt for model change/download).
- Trimmed app package size by -6.64 MB (over half previous release).
Wear OS app (359110304W)
- Introduces local audio recording support:
- Uses a foreground service (additional permission used) to maintain access to the microphone regardless of whether the application is open or moved to the background.
- The home screen on the Watch will show a microphone symbol indicating that recording is active.
- Introduce a status toolbar:
- Shows device (Android) connected state.
- Queued message count; when the Watch is disconnected from the device messages/audio files are stored in a queue to be synced upon reconnection.
- Recognition type: whether Google STT or Audio recording is selected.
- Recording indicator: only visible when Audio recording is enabled and active.
- Introduces support to cache messages/audio offline this enables the Wear OS application to operate more independently from requiring an active connection to the Android app.
- Note that Google Speech-to-text still requires an active internet connection; this could be via Wifi or via a Data connection if you watch supports/has its own data plan.
- Audio recording operates offline; the only requirement to be able to process audio speech-to-text is that the Transcribable Whisper service is configured on the Android device.
- The Android app will notify the Wear OS app is this is not configured.
- Recordings can still be made even when the the Transcribable Whisper service is not yet configured.
- Adds a Queue activity access via swiping down to the top clock area:
- Lists pending messages/audio to be synced to the device.
- Adds a Settings activity to tweak the Wear OS application behaviour:
- Switch between Google Speech-to-text or Audio recording.
- Changes the clock format between 12hr/24hr.
- Use the Watches back button as the Start/Stop recognition/recording trigger.
- Hide the home screen help content.
- Dates shown in the application and sent by the app now uses the devices locale settings.
- Improvements to Ambient mode:
- Status toolbar goes in to a dimmed state without colour; several other elements are hidden in Ambient mode.
- Improvements to Android device connectivity checks.
- Data Syncing between the watch and Android device uses the DataLayer service and operates locally (via bluetooth or Wifi).
- If the watch has been disconnected from the device and has queued items to be synced. This will trigger within a few minutes of reconnecting to the Android.
- Note: When Wear OS power saving is enabled this may delay transmitting data immediately but it will sync.
Learn more about Transcribable.
