Local-first voice note transcription

Turn voice notes into a Markdown inbox.

Voice Inbox turns accumulated recordings into a readable text inbox. Android has the mature folder workflow today; iOS now has an active MVP for importing, previewing, transcribing, and appending individual recordings locally.

Voice Inbox Android app showing scanned audio files and playback actions

Android workflow screenshot. iOS MVP is evolving alongside the shared core.

Platform status

Android is ready; iOS is catching up fast.

Android ready

Full folder workflow

Scan a selected audio folder, preview recordings, transcribe individual files or all New files, append to Markdown or text, retry failures, and schedule best-effort background runs.

iOS MVP

Manual import workflow

Import audio files, preview them, download the speech model, transcribe locally, show recognized text, and append output. Folder automation and scheduling are still in progress.

Shared core

Kotlin Multiplatform foundation

Catalog and transcription use cases are moving into shared code so Android and iOS can converge without hiding platform-specific storage and background limitations.

What it does

Built for accumulated recordings, not live dictation.

Incremental folder transcription

Select one folder, review discovered recordings, and process only files that are new or changed.

Local recognition

After the first model download, recognition runs on device instead of sending audio to an app cloud service.

Markdown or plain text output

Each successful note is appended to an existing writable .md or .txt document.

Preview before processing

Play audio files from the list before deciding to transcribe them.

Scheduled processing

On Android, optionally ask the app to run scheduled transcription around your chosen time. Exact execution is not guaranteed.

Transparent history

Processed and failed files stay visible, and failed files can be retried individually.

Workflow

From recordings to notes in five steps.

Install the model

Download and verify the speech model on first launch.

Select output

Choose an existing writable Markdown or text file.

Select recordings

Pick the folder that contains your audio notes.

Review New files

Preview recordings and check what will be processed.

Transcribe all

Append one entry per successful file with filename, recording time, and recognized text.

Requirements

Designed for local, on-device recognition.

Android requires Android 7.0 (API 24) or later on an ARM64 device. Both Android and the iOS MVP need network access for the first model download and enough free storage for model installation.