opensourceprojects.dev

A broadsheet for software that doesn't ask for your email

WhisperBoard brings offline Whisper transcription to iOS with one-tap recording
GitHub RepoImpressions14

Project Description

View on GitHub

WhisperBoard: Offline Speech-to-Text on iOS, No Cloud Required

Transcribing your voice notes without ever sending audio to a server.

If you've ever wished Siri's dictation worked offline, or wanted to transcribe long meetings without worrying about data caps or privacy, WhisperBoard is exactly what you're looking for. It's an iOS app that lets you record audio and get instant transcriptions using OpenAI's Whisper model—completely on-device.

No internet connection. No uploads. Just your microphone, your voice, and a local AI.


What It Does

WhisperBoard is a simple, minimal iOS app. You tap a button, speak, and it transcribes your speech into text in real time. The entire pipeline runs locally on your iPhone or iPad, using Apple's Core ML optimizations to make the Whisper model fast enough for offline use.

It's effectively a voice notes app with state-of-the-art transcription built in, but without the usual "requires internet" asterisk.


Why It's Cool

Offline-first transcription. This is the big one. Most transcription apps (including OpenAI's own Whisper API) require sending your audio to a remote server. WhisperBoard keeps everything local. Your voice never leaves your device. For anyone privacy-conscious, or developers working with sensitive audio, this is a game changer.

One tap recording. No setup, no multiple steps. Open the app, tap record, start speaking. Tap stop. Transcription appears instantly. It feels like a native iOS feature, not a third-party app.

Uses the real Whisper model. This isn't a lightweight, less accurate alternative. It's the actual Whisper model (base or small, depending on your device), compiled to run on the Neural Engine. The quality is genuinely impressive for local inference.

Developer-friendly. The repo is clean, well-structured, and built with SwiftUI. If you want to understand how to integrate Whisper into an iOS app, or how to optimize Core ML models for on-device speech recognition, WhisperBoard is excellent reference code. It also shows how to handle model downloading and caching gracefully.


How to Try It

  1. Go to the GitHub repo: https://github.com/Saik0s/Whisperboard
  2. Clone or download the project.
  3. Open Whisperboard.xcodeproj in Xcode.
  4. Build and run on a physical iPhone or iPad (simulator won't work for microphone access or Neural Engine inference).
  5. The app will prompt you to download the Whisper model on first launch. This happens once.
  6. Tap record, talk, tap stop, and see your transcription.

That's it. No third-party keys, no accounts, no API limits.


Final Thoughts

WhisperBoard is one of those projects that makes you wonder why the official iOS dictation doesn't work this well offline. For developers, it's a great sandbox for playing with on-device AI. For anyone who records meetings, lectures, or voice notes, it's a genuinely useful tool that respects your privacy.

It's not flashy. It just works. And that's a rare thing in the world of AI-powered apps.


Follow us on X: @githubprojects

Back to Projects
Project ID: 22a086de-1de7-40e0-976c-42a7a518c2daLast updated: July 19, 2026 at 02:44 AM