Pindrop is a native macOS menu bar dictation application designed for users who prioritize privacy, performance, and a seamless user experience. It addresses the need for a fully offline, local speech-to-text solution, leveraging advanced AI models without compromising data security or requiring cloud connectivity. The application offers a significant benefit by providing fast, accurate transcriptions optimized for Apple Silicon hardware.

How It Works

Pindrop is built entirely with Swift and SwiftUI, ensuring a true native macOS feel. Its core functionality relies on WhisperKit, an optimized Core ML implementation of OpenAI's Whisper model, enabling 100% local transcription. This approach is advantageous as it guarantees user privacy by keeping all audio and transcription data on the user's machine. Furthermore, WhisperKit's specific optimizations for Apple Silicon processors deliver transcription speeds 2-3 times faster than generic Whisper implementations, minimizing battery drain and maximizing responsiveness.

Quick Start & Requirements

• Primary Install: Download the Pindrop.dmg from the releases page and drag the application to the /Applications folder. Due to being self-signed, the first launch requires a right-click -> "Open" action, followed by granting developer verification in System Settings > Privacy & Security.
• Prerequisites: macOS 14.0 (Sonoma) or later. Apple Silicon (M1/M2/M3/M4) is highly recommended for optimal performance. Microphone access is mandatory; Accessibility permission is optional for direct text insertion.
• Links: Latest Release, Documentation, Contributing, Community.
• Building from Source: Clone the repository and open Pindrop.xcodeproj in Xcode, or use the provided justfile with commands like just build-release or just dmg-self-signed.

Highlighted Details

• 100% local transcription ensures complete data privacy.
• Supports multiple transcription engines, including WhisperKit (Core ML) and Parakeet, with streaming capabilities.
• Global hotkeys for configurable toggle or push-to-talk functionality.
• Smart output automatically copies text to the clipboard and can optionally insert it directly into applications.
• Integrated notes system with pinning, tagging, and AI-powered title generation.
• Comprehensive transcription history with local storage, search, and export options (JSON, CSV, plain text).
• Optional AI enhancement for transcription cleanup via any OpenAI-compatible API.
• Custom dictionary support for specialized vocabulary and word replacements.
• Automatic updates managed by the Sparkle framework.

Maintenance & Community

The project encourages community involvement through GitHub Discussions for questions and ideas, and GitHub Issues for bug reports and feature requests. Contribution guidelines are available in CONTRIBUTING.md.

Licensing & Compatibility

Pindrop is released under the MIT License, which is highly permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

The application is exclusively designed for macOS. While Intel Macs are supported, optimal performance and efficiency are achieved on Apple Silicon hardware. The current distribution method involves self-signed builds, necessitating manual user approval on first launch due to the absence of an Apple Developer ID.