Velora

Velora is a local-first dictation tool designed specifically for macOS on Apple Silicon. This innovative application transforms voice input into text efficiently while ensuring complete privacy. Users simply hold a key, speak, and release to see polished text appear in any app being used. With every dictation step operating on-device via MLX, functionality remains completely offline, ensuring audio, transcripts, and history never leave the user's Mac. Confirmed Personal Dictionary terms can sync privately through iCloud Drive.

Why Choose Velora?

Existing dictation tools like Superwhisper and Wispr Flow may provide certain conveniences, yet they process data in the cloud and often require a subscription. Velora redefines this approach with an open-source alternative that emphasizes privacy and speed:

• Privacy by Design: No network calls at dictation time; models are downloaded only once from Hugging Face.
• Open-Source and Free: Velora is MIT-licensed, granting ownership of modes, prompts, and vocabulary.
• Fast Performance: Speech-to-text processing occurs in real-time while users talk, ensuring that polished text appears almost instantaneously upon releasing the key.
• Trusted Clean-Up: AI cleanup features are conservatively designed to avoid unwanted alterations, ensuring users maintain control of their content.

Key Features

• Flexible Dictation Control: Users can dictate by holding the Right-Option key, toggling dictation mode with a double-tap or menu icon, all of which are configurable.
• Capsule HUD: A discreet floating capsule displays a live waveform of audio input during dictation, transitioning smoothly through various states.
• Multilingual Support: The default whisper-large-v3-turbo model adeptly handles multiple languages including English, Hindi, and others, with automatic language detection.
• Smart Formatting Based on Apps: Velora integrates intelligently with the active application to apply appropriate formatting, adapting according to the context (e.g., casual for messaging apps or structured for email).
• Local Speech-to-Text and AI Clean-Up: Transcription and cleanup both occur on-device, with no reliance on internet connectivity.
• History Browser: A local database maintains dictations that can be browsed, searched, and reused.
• Voice Intelligence Dashboard: Users can explore transcription trends, estimated time saved, and other insights without compromising privacy.
• Private Meeting Memory: The tool can facilitate recording during meetings, ensuring that all data remains local and requires explicit user confirmation.
• Safe Voice Editing: Users can modify existing text directly by selecting it and dictating instructions for changes, with robust safety measures against undesired alterations.
• Easy Customization: Users can create and modify modes and prompts through simple configuration files to tailor their dictation experience.

Technical Details

• Requires Apple Silicon Mac (M1 or later) and macOS 14+.
• Dependencies include uv for managing the Python engine.
• The installation includes steps for building from source or using Homebrew for ease.

With its commitment to local processing, user privacy, and customizable features, Velora stands out as a powerful dictation solution tailored for modern users demanding speed and security in their voice-to-text applications.