Skip to content

Getting Started

Requirements

Requirement Details
Python 3.13 or newer
Microphone Any input device recognised by your OS
NVIDIA GPU Required for GPU-accelerated transcription. See GPU Acceleration
uv Required for installation. Install uv

NVIDIA GPUs only

GPU acceleration has been tested exclusively with NVIDIA GPUs (RTX 30/40 series). AMD and Apple Silicon GPUs are not supported. VoicePad can run in CPU-only mode, but transcription will be significantly slower.

Install uv

VoicePad is installed and run using uv, the fast Python package manager from Astral.

winget install --id=astral-sh.uv -e

curl -LsSf https://astral.sh/uv/install.sh | sh

After installing, restart your terminal so the uv and uvx commands are available.

Installation

uvx runs VoicePad in a temporary, isolated environment. Nothing is permanently installed on your system.

uvx voicepad

On first run, it downloads the package and the default Whisper model (turbo, ~800 MB). Subsequent runs start immediately from cache.

Option 2: Install as a persistent tool

If you want VoicePad permanently available in your shell:

uv tool install voicepad
voicepad

This installs VoicePad into an isolated environment managed by uv and adds the voicepad command to your PATH.

First Run

When you run VoicePad for the first time:

  1. The interface opens and shows initialising in the header
  2. The turbo Whisper model downloads from HuggingFace (~800 MB, one time only)
  3. The model loads into GPU memory (or CPU if no NVIDIA GPU is detected)
  4. The status changes to ready. You can now record

Subsequent launches are instant

The model is cached locally after the first download. Future launches skip the download and start in seconds.

Your First Recording

  1. Open VoicePad: uvx voicepad
  2. Make sure you are on the Record tab (selected by default)
  3. Press Space to start recording. The status changes to recording
  4. Speak clearly into your microphone
  5. Press Space again to stop. Transcription begins immediately
  6. The transcribed text appears in the transcription panel within seconds

The recording is saved as a WAV file and the transcription as a markdown file. Both go to ~/.config/voicepad/data/ by default. See Configuration to change this.

Check Your Microphone

If nothing is captured, your microphone may not be set as the default input device. You can change the input device from the Settings tab in VoicePad.

Next Steps