Skip to content

Installation

git clone https://github.com/luckeyfaraday/athena-whisper.git
cd athena-whisper
python3 -m venv .venv
. .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -e ".[dev,gui]"

Linux — for X11 desktop insertion:

sudo apt-get install xdotool xclip

If PyQt6 reports an XCB platform plugin error, install the common X11 Qt runtime libraries:

sudo apt-get install libxcb-cursor0 libxcb-xinerama0 libxkbcommon-x11-0

Windows — no additional tools required. The windows-keystrokes and windows-clipboard-paste backends use ctypes to call Win32 SendInput and keybd_event directly.

.venv/bin/athena-dictate widget # Linux
.venv\Scripts\athena-dictate widget # Windows

The repository includes an athena-dictate.toml configured for automatic platform-specific insertion. On Linux/X11, auto uses keystrokes instead of clipboard paste, so terminal apps such as Codex, Claude Code, opencode, and shell-based TUIs do not receive Ctrl+V image-paste shortcuts. On Windows, auto uses Win32 keystrokes.

Check the current desktop/session environment:

athena-dictate doctor

Transcribe an existing audio file:

athena-dictate transcribe-file path/to/audio.wav
athena-dictate transcribe-file path/to/audio.wav --language es # explicit language
athena-dictate transcribe-file path/to/audio.wav --language auto # auto-detect
athena-dictate transcribe-file path/to/audio.wav --task translate # translate to English

Record one short dictation and insert the result:

athena-dictate record-once --seconds 5 --paste
athena-dictate record-once --multilingual # language changes within one recording
athena-dictate record-once --seconds 5 --no-paste # record without inserting

Test keystroke insertion without recording or transcribing:

athena-dictate type-text "hello from athena"
.venv/bin/pytest
.venv/bin/python -m compileall src tests

The current repository does not include PyInstaller packaging scripts or released desktop installers. Run Athena Whisper from the source checkout for now.