Velox - AI voice input

Introduction to Velox - AI voice input

Velox AI Input is a macOS application designed to streamline voice-to-text transcription for users who rely on spoken input during their daily workflows. Built exclusively for macOS Sonoma and later, it replaces traditional dictation systems with a seamless, focused experience that integrates directly into active applications. By using a single key press, users can speak naturally, receive polished text output, and have it pasted automatically—without switching windows or managing separate interfaces.

The application targets professionals, writers, developers, and anyone engaged in frequent text entry across messaging apps, documents, email, and terminal environments. Its design philosophy emphasizes minimalism, privacy, and efficiency, ensuring the tool remains unobtrusive while delivering reliable performance tailored to modern writing needs.

Key Takeaways

• Press and hold the Option (⌥) key to record speech; release to transcribe, refine, and paste
• Supports four speech-to-text providers: OpenAI, Groq, Alibaba Cloud, and custom Whisper-compatible endpoints
• Operates with no proprietary servers—users configure their own API keys and services
• Maintains clipboard integrity by saving and restoring content before pasting
• Features smart focus awareness to prevent accidental pasting into incorrect applications
• Offers light text correction including punctuation and basic grammar improvements
• Interface supports seven languages; UI language and output language can be set independently
• Local storage of audio, history, and credentials—no user data sent to external servers

How Velox AI Input Works

The core interaction begins by pressing and holding the Option (⌥) key, which activates a floating heads-up display (HUD) near the bottom of the screen. Recording starts immediately, capturing audio through the system microphone. When the user releases the key, recording stops, and the audio is processed using the configured speech-to-text provider.

The resulting text undergoes light refinement to improve clarity, add proper punctuation, and correct minor speech artifacts—all while preserving the user’s original intent and tone. If the active application remains in focus, the refined text is pasted directly via automated keyboard simulation (using accessibility permissions). If the user switches apps during processing, the result is cached in the HUD for manual copying when ready. Audio files are stored temporarily within the app sandbox and deleted after each session.

Core Benefits and Applications

Velox AI Input enhances productivity in scenarios where rapid, hands-free text input is valuable. It is particularly useful in continuous writing tasks such as composing emails, drafting documentation, coding comments, or participating in real-time chat conversations. The ability to work without launching auxiliary windows reduces cognitive load and maintains workflow continuity.

Privacy is central to its architecture: all sensitive data, including API keys, are stored in macOS Keychain, and no content from user applications is recorded or transmitted. Users retain full control over cost and data by bringing their own API credentials—whether connecting to cloud-based services like Groq or running self-hosted Whisper-compatible models locally.

The dual-language system allows users to operate with an English interface while generating output in another language, such as Traditional Chinese or Japanese, supporting multilingual workflows. With local history, adjustable HUD positioning, and compatibility across Apple Silicon and Intel Macs, Velox provides a robust, transparent solution for secure and efficient voice input on macOS.

System Requirements

• macOS 14.6 (Sonoma) or later
• Microphone access
• Accessibility permission for paste automation
• 8 GB RAM recommended for heavy use
• Stable internet connection (when using cloud-based STT providers)