OpenScribe

Native macOS transcription app scaffold built for WhisperKit-based workflows.

Features in this build

• Microphone recording and transcription flow
• File upload/transcription flow
• System-audio capture flow entry point
• On-demand model download with local cache
• Model manager (download/delete/size)
• Download cancel + automatic retry
• Model picker and speaker detection toggle
• Segment view in UI
• TXT export
• Live recording indicator with level meter

Requirements

• macOS 14+
• Xcode 15.4+ or Swift 5.10+
• No external CLI required (direct WhisperKit + SpeakerKit integration)

Run

swift run OpenScribe

Notes

• Models are downloaded via WhisperKit directly and cached under: ~/Library/Application Support/OpenScribe/ModelCache
• Download jobs support progress, cancelation, and retry.
• Optional checksum verification is supported via checksumSHA256 in Sources/OpenScribe/Resources/models.json.
• System audio capture uses ScreenCaptureKit and requires Screen Recording permission in macOS Privacy settings.