VoicePad

Your private, local-first dictation studio.

Record audio and get instant transcriptions — entirely on your machine. No cloud, no subscriptions, no data leaving your hardware.

uvx voicepad

Why VoicePad?

Most transcription tools ship your voice to a cloud API. VoicePad doesn't. Everything runs locally, powered by faster-whisper and your NVIDIA GPU.

• 100% local — audio never leaves your machine
• GPU-accelerated — 60-second recording transcribed in ~3 seconds on an RTX 3050
• Re-transcribe anytime — original WAV files are always saved alongside your markdown notes
• Zero setup — CUDA libraries are bundled, no separate CUDA install needed

Quick Start

Requires uv.

# Run instantly, no install needed
uvx voicepad

# Or install permanently
uv tool install voicepad
voicepad

Press Space to record, Space again to stop. Transcription starts immediately.

Requirements

• Python 3.13+
• Windows or Linux
• Any microphone your OS recognises
• NVIDIA GPU (optional, but recommended for fast transcription)

Documentation

Full docs at voicepad.hyperoot.dev

• Getting Started
• Interface
• Configuration
• GPU Acceleration

License

MIT