# Install

The GitHub setup needs no install. Install the command-line tool only if you want to dictate on your own computer, for example to edit text before it is dictated, or to host your library somewhere other than GitHub.

You need [uv](https://docs.astral.sh/uv/) and `ffmpeg`. Install `ffmpeg` for your platform:


- <a href="" id="tabset-1-1-tab" class="nav-link active" data-bs-toggle="tab" data-bs-target="#tabset-1-1" role="tab" aria-controls="tabset-1-1" aria-selected="true">macOS</a>
- <a href="" id="tabset-1-2-tab" class="nav-link" data-bs-toggle="tab" data-bs-target="#tabset-1-2" role="tab" aria-controls="tabset-1-2" aria-selected="false">Linux</a>
- <a href="" id="tabset-1-3-tab" class="nav-link" data-bs-toggle="tab" data-bs-target="#tabset-1-3" role="tab" aria-controls="tabset-1-3" aria-selected="false">Windows</a>


``` bash
brew install ffmpeg
```


``` bash
sudo apt install ffmpeg     # Debian, Ubuntu
sudo dnf install ffmpeg     # Fedora
```


``` powershell
winget install Gyan.FFmpeg
```

Open a new terminal afterwards so that `ffmpeg` is on your `PATH`. Windows is expected to work, but it is not tested.


Then install earmark:

``` bash
uv tool install --python 3.12 git+https://github.com/earmark-dev/earmark
earmark --version
```

`--python 3.12` is necessary. The speech model does not support Python 3.13 or newer, so uv downloads Python 3.12 for earmark.

The first time you dictate something, earmark asks to download the 354 MB speech model. This happens once. The model is stored with your application data, not in your library.

Next: [Use the CLI](use.md).