Setup guide
Get Voice Inbox ready for your recordings.
Voice Inbox writes recognized speech to a user-selected output document and processes audio that you import, share, or discover in an optional folder. Android has the mature task-list workflow today; iOS is an active MVP. Model installation is large, but recognition runs locally after it completes.
1. Install the app
Voice Inbox is currently available through a restricted Google Play closed test. Follow the closed-testing instructions to join the tester group, opt in with the same Google account, and install the app on an ARM64 Android device running Android 7.0 or later.
Android source builds and APKs remain available from GitHub Releases. iOS is currently an MVP built from source rather than an App Store release.
2. Download the speech model
Download and verify the pinned int8 Parakeet model from the model task. The download is about 640 MiB, and installation needs about 704 MiB of free storage. If the download source is unavailable, Android can instead install the same model from a compatible local folder.
Network access is required for the normal download, but not for a valid local-folder installation. Recognition runs locally after the model is installed.
3. Select an output document
Use Create New to create a writable .txt or .md document, or
Choose Existing to select one you already use. Voice Inbox appends new entries and
preserves existing content. On iOS, output selection uses the document picker and security-scoped access.
4. Add recordings
On Android, use Import Audio Files for one or more recordings, or share an audio file to Voice Inbox from another app. Selecting an audio folder is optional and useful for batch processing; the app scans direct children only and does not traverse nested folders.
On iOS, import app-local files, receive audio through the Share Extension, or select a folder to refresh its direct audio files into the queue.
Optional: dictate with the Android voice keyboard
Android can use the installed local speech model for short-phrase dictation in another app. The keyboard needs microphone permission and a ready model; it is separate from importing or scanning recordings into the Voice Inbox task list.
5. Review and transcribe
The main screen presents setup and audio work as a task list. Use Play to preview a recording, transcribe it individually, or choose Transcribe All when multiple tasks are ready.
Failed files move to the Processed tab with their latest error and a Retry action. Successful files are appended to the output document.
6. Optional: startup processing
In Settings, choose whether Voice Inbox should ask, transcribe automatically, or leave files queued when the app opens and finishes scanning. Ask is the default. Startup processing is available on Android and iOS.
7. Optional: nightly transcription
On Android, open Settings to enable nightly transcription and choose a preferred time. Android may delay background work, so scheduled runs should be treated as best-effort rather than exact alarms.
Nightly background transcription is independent from startup processing and is not part of the current iOS MVP.