Phantom Voice

Product Introduction

1. Definition: Phantom Voice is a macOS-native, system-wide push-to-talk dictation utility. It is a local transcription tool that leverages the open-source whisper.cpp speech-to-text engine accelerated by Apple's Metal API for on-device processing.
2. Core Value Proposition: It exists to provide a private, fast, and seamless dictation experience for macOS users, eliminating the need to send audio data to the cloud. Its primary value is local, offline transcription with a one-time purchase model, directly inserting formatted text into any active application.

Main Features

1. Local whisper.cpp Transcription: Audio is processed entirely on the user's Mac. The speech recognition model (e.g., Whisper Large v3 Turbo) is downloaded once and runs locally using Metal acceleration on Apple silicon, ensuring zero data transmission to external servers. This guarantees privacy and enables offline functionality.
2. System-Wide Push-to-Talk Insertion: The core interaction is a hardware-level hotkey (default: fn). The user holds the key to speak and releases to process. The finished text is inserted directly at the cursor's location in the currently focused app—be it a code editor, terminal, document, or messaging client—without requiring app switching or copy-paste.
3. Customizable, Animated HUD Indicators: Unlike generic recording dots, Phantom Voice features hand-animated visual indicators for its states (Ready, Listening, Processing). Users can choose from multiple visual styles (like White Wave, Retro Fox) and sizes. These are not just skins but distinct, frame-by-frame animations that provide clear, aesthetic feedback.
4. Optional On-Device Cleanup Pass: An additional, entirely local AI model (~1.1 GB) can be enabled to perform post-transcription cleanup. This refines punctuation, removes filler words (e.g., "um," "ah"), and applies a personal dictionary without ever sending the text to an external API or service.
5. Perpetual License with Trial: The software operates on a one-time payment of $10 with no subscription. It includes a full-featured 14-day free trial. Additional visual indicator styles (Retro Dog, Real Revolver, Piano Keys) are available as separate, one-time $3 add-ons.

Problems Solved

1. Pain Point: Privacy concerns with cloud-based dictation services where audio recordings and transcripts are processed on external servers, creating potential data security and compliance risks.
2. Pain Point: Disruptive workflow of traditional dictation tools that often require switching to a separate app, manually copying transcribed text, and pasting it into the target application.
3. Target Audience: Software developers and engineers writing code, documentation, or commit messages; writers and journalists drafting content; knowledge workers and students taking notes or composing emails; accessibility-focused users seeking efficient hands-free input; privacy-conscious professionals in legal, healthcare, or business sectors.
4. Use Cases: Dictating code snippets and comments directly into an IDE like VS Code or Xcode; writing long-form documentation or emails without leaving the text field; providing verbose CLI commands in Terminal; quickly logging issues in project management tools; composing messages in Slack or Discord while in a full-screen application.

Unique Advantages

1. Differentiation vs. Cloud Services: Unlike Siri, Google Dictate, or Otter.ai, Phantom Voice performs 100% local transcription, ensuring audio never leaves the device. It contrasts with subscription-based models by offering a single, upfront payment.
2. Differentiation vs. Other Local Tools: Compared to other whisper.cpp front-ends, its deep macOS integration (Accessibility/Input Monitoring for system-wide hotkey), polished native UI with custom animations, and the unique "insert-in-place" workflow provide a significantly more seamless and user-friendly experience.
3. Key Innovation: The combination of low-level system integration for reliable push-to-talk across all apps, a highly optimized local inference pipeline (whisper.cpp + Metal), and a focus on deliberate, artistic user feedback (the HUD indicators) creates a product that is both technically robust and thoughtfully designed.

Frequently Asked Questions (FAQ)

1. Does Phantom Voice work offline? Yes, after the initial download of the speech recognition model, Phantom Voice requires no internet connection for dictation. All audio processing and transcription occur locally on your Mac.
2. What are the system requirements for Phantom Voice? It requires macOS 14 (Sonoma) or later on a Mac with Apple silicon (M-series chips). It transcribes English and requires microphone, accessibility, and input monitoring permissions to function system-wide.
3. How does the optional local cleanup work? When enabled, a separate on-device AI model analyzes the initial whisper.cpp transcript. It improves punctuation, removes verbal filler words, and applies corrections based on a local dictionary, all processed locally without any data being sent to the cloud.
4. Can I use Phantom Voice in any application? Yes, it is designed to work system-wide. As long as an application has a standard text input field (code editors, word processors, browsers, terminals, chat apps), Phantom Voice can insert transcribed text into it upon releasing the push-to-talk key.
5. What is the difference between the included and paid HUD styles? Phantom Voice includes two indicator styles (White Wave, Retro Fox) with the base purchase. The paid add-ons (Retro Dog, Real Revolver, Piano Keys) are additional, distinct animated styles offered as separate one-time purchases. All styles function identically; the difference is purely visual.