Improved voice typing with the Offline Voice Input (Android) app

Lounge
1

I'm still having the same issue I reported last year in Voice typing with Whisper keeps adding "It's a note-taking application." to my note - Support - Joplin Forum. In addition, too many transcriptions get cut off after an arbitrary amount of time.

Having successfully used Handy for voice transcription on desktop, I wanted to see if there was an Android equivalent. I checked Handy's Github issues for any mentioning "Android". This issue, [ Question ] Any plans to make android app? · Issue #171 · cjpais/Handy, spawned this discussion, Android app? · cjpais/Handy · Discussion #195 · GitHub, where someone announced their take on Handy for Android late last year:

I created an android app using transcribe-rs (the library of Handy) and by creating an IME.

It's an IME (Input Method Editor) like Sayboard or FUTO Voice Input, which means it's usable beyond Joplin.

The Handy maintainer responded favorably to this news. The original issue submitter said this about the app in the above discussion:

Wow, just tested it on my older mid-ranger (5yo device) and it works really well!

Having tested it myself, it solves the pain points I have with Joplin's voice typing feature. With this topic having been discussed before at Voice memos to Joplin text notes - Features - Joplin Forum, I would like to share this app with you. You can find and download it at these locations:

Read on for my first impressions of v0.1.14 of the app:

Features:

  • Free and open source, offline, and privacy-focused unlike popular Android transcription options such as WisprFlow, Typeless, etc.
  • Uses Nvidia's Parakeet speech-to-text model and reuses Handy's inference backend, transcribe-rs.
  • In my experience, it processes longer recordings than Joplin's voice typing feature with fewer errors and no canned phrases.
  • Supports 25 languages so far.
  • Keeps the editor cursor visible unlike Joplin's voice typing feature.
  • Recording doesn’t appear to be time-limited nor prevents app interactions like the last time I tried FUTO Voice Input.

Notes:

  • Transcription isn't on-the-go like voice typing. Processing starts once the user stops the recording. Text is pasted wherever the cursor is, even if it was moved during recording/processing, once processing finishes.
  • Only has Space, Enter, and Backspace buttons so far compared to Sayboard's larger button set.
  • Pasted text isn’t split into paragraphs. For line breaks, either make multiple recordings or use the Enter button as needed after text is transcribed.
  • Only tapping Tap to Stop will start processing your recording.
  • Lacks customization or custom words like FUTO Voice Input, but I don't use features like these myself.
  • Has a live subtitles feature for real-time captions for on-device audio/video.

Hope this write-up helps for anyone looking for a solid, free, and private voice transcription Android app for use with Joplin or otherwise.

1 Like