What it does

Whisperboard is an iOS voice transcription app built around OpenAI’s Whisper model via whisper.cpp. Record audio, import files, pick your microphone, and get text back. It also lets you browse and download different Whisper models directly in the app to trade off accuracy against speed and size.

The interesting bit

The resumable transcription is the quietly useful part: kill the app mid-transcription and it picks up where it left off. That’s the kind of phone-reality feature more apps should steal. The roadmap also flags real-time transcription with smaller models, though that’s not shipped yet.

Key highlights

• Runs Whisper locally on iOS through whisper.cpp — no network required for transcription
• Built in SwiftUI with The Composable Architecture (TCA) and Tuist for project generation
• GPL-3.0 licensed, so fork away
• One-tap recording, audio import/export, and selectable microphones
• Model browser lets you download different Whisper variants to suit your hardware

Caveats

• Real-time transcription is on the roadmap but not implemented; current transcription is batch-style after recording
• The README’s “unparalleled accuracy” claim is marketing fluff — actual accuracy depends on your model choice and audio quality
• Requires building from source with make and Xcode; no TestFlight or App Store link mentioned

Verdict

Worth a look if you want a hackable, offline-first transcription app on iOS or need a reference for running whisper.cpp in a SwiftUI/TCA project. Skip it if you need real-time results today or want something you can install without Xcode.