Voice Doom

Voice Doom was a hack at alt+ctrl Mod Jam #03 in Newspeak House, London on 28th March 2026. The idea was to control Doom with you voice. See video of it in action. It is made of two components:

1. Sound input process

A modified graphic equalizer demo program based on the original https://github.com/spatialaudio/python-sounddevice/blob/master/examples/spectrog... from the sounddevice Python package https://pypi.org/project/sounddevice/.

This ran in its own process, parsing the spectogram to extract different patterns of frequency and volume. See modified source code at https://gitlab.com/-/snippets/5976881 the big "FRANCIS WAS HERE". It writes the direction to press in /tmp/doom.key (as a very easy/cheap inter-process communication)

2. Modified Doom

The easiest way I found to be able to edit the core input code around DOOM, was using doomgeneric, and altering its SDL implementation https://github.com/ozkl/doomgeneric/blob/master/doomgeneric/doomgeneric_sdl.c

Look at the handleKeyInput in the source code at https://gitlab.com/-/snippets/5976881 for the hacky code that reads the keystroke file and modifies keyboard inputs.

Thoughts

It was relatively easy to detect pitch and volume, but even with years of choir practice, quite hard when stressed to get the right notes. Loud overall volume (i.e. spread out across spectrum more like a crackly white noise) was great for shooting. And had various combinations of pitches for turning and moving.

It was really lovely to be able to hold a note and keep doing something, and stop exactly. It felt really responsive, like you could control a whole game this way, if it was designed for it.