[add] Speech to text feature.

README.md

@@ -31,6 +31,65 @@ You can also run it directly as a module without installing:
 uv run python -m h2g2.main
 ```
 
+## Audio features
+
+### Text-to-speech (TTS)
+
+The game can read all output aloud using [Piper](https://github.com/rhasspy/piper), a fast offline TTS engine. A British English voice model is included in the repo.
+
+```bash
+uv run h2g2 --audio
+```
+
+Use `--voice /path/to/model.onnx` to use a different Piper voice model.
+
+### Speech-to-text (STT)
+
+You can play the game hands-free using voice input powered by [Vosk](https://alphacephei.com/vosk/), a lightweight offline speech recognition engine. The Vosk model (~50 MB) downloads automatically on first use.
+
+```bash
+uv run h2g2 --stt
+```
+
+Use `--stt-model /path/to/vosk-model/` to use a different Vosk model.
+
+Combine both flags for full voice interaction:
+
+```bash
+uv run h2g2 --audio --stt
+```
+
+### Audio prerequisites
+
+STT requires PortAudio for microphone access. Install the system library for your platform before running with `--stt`:
+
+**Linux (Debian/Ubuntu):**
+
+```bash
+sudo apt install libportaudio2 portaudio19-dev
+```
+
+**macOS:**
+
+```bash
+brew install portaudio
+```
+
+**Windows:**
+
+PyAudio ships with PortAudio bundled on Windows, so no extra system package is needed. If you run into build issues, install a prebuilt wheel:
+
+```bash
+uv pip install pipwin
+pipwin install pyaudio
+```
+
+After installing the system dependency, sync the Python packages:
+
+```bash
+uv sync
+```
+
 ## What's playable
 
 The Earth opening sequence: wake up in your bedroom, find your gown and aspirin, make your way downstairs, head to the pub, and meet Ford Prefect.