fossasia/voxbento-local

What it solves

Voxa solves the privacy and security risks associated with cloud-based meeting assistants. It allows users to record, transcribe, and summarize meetings entirely on their own computer, ensuring that sensitive audio data never leaves the local machine or touches third-party servers.

How it works

Built using the Tauri framework, Voxa operates as a desktop application that captures both microphone input and system audio (such as Zoom or Teams calls) directly. It uses local AI models like Whisper and NVIDIA Parakeet to transcribe speech and leverages Ollama or embedded tools to generate summaries. All data, including transcripts and audio files, is stored in a local SQLite database, and the application provides a workspace to review, edit, and export meeting notes in Markdown or plain text.

Who it’s for

Voxa is designed for privacy-conscious professionals, researchers, and teams who need to document meetings and generate action items without relying on external cloud services or intrusive meeting bots.

Highlights

• 100% On-Device AI: Uses hardware-accelerated local models for speech-to-text.
• Zero Cloud Audio Upload: Keeps all raw recordings and audio streams strictly on your physical machine.
• Dual-Channel Capture: Records both your microphone and system audio without requiring meeting permissions or external bots.
• Local Summarization: Generates executive summaries and action items using local LLMs.
• Structured Workspace: Includes timestamped transcripts, synchronized audio playback, and easy export options.