getting my computer to play the windows 95 startup sound

Yesterday i had what i called a “profoundly terrible idea”. What is that terrible idea, you may ask? I’ll start with the payoff:

last night i had a profoundly stupid idea, and today i got it working

(embedded video)

~@QuietMisdreavus, 2020-06-07 11:19 AM -06:00

I wanted my computer to play the Windows 95 startup sound when it booted up, like i was back in the 90’s and playing sound when the computer was ready was cool and expected. Mostly because i thought it was funny and it seemed like something i could configure relatively simply.

Some background: My desktop currently runs Arch Linux, which i’ve been customizing over the last week or so for fun. I’ve used Arch a lot in the past, but i hadn’t used it as a “daily driver” desktop OS for a while, and i wanted to try out GNOME 3 as a desktop environment. I’d recently set up a common-use computer with Ubuntu for my roommates and i thought that it would be nice to use that kind of environment for my personal computer. I have it in a spot where i like it, and over the last few days i’ve just been getting more and more personal with the tweaks i’ve been putting in.

Which brings us to today’s post. I could find a sample of the Windows 95 startup chime fairly easily. But after that, the trick became how to wedge that into the startup process of my computer. I knew that the first thing that my computer ran when it could get a graphical environment running was GDM, so i started looking at that Arch Wiki article to see what could be done to customize it. But the methods it had for customizing it didn’t mention anything about playing things at startup, and digging around the method it mentioned (logging in as the GDM user and setting configuration options directly) also turned up nothing.

On a lark, i looked at the file listing for the GDM package on Arch, and found a folder named /usr/share/gdm/greeter/autostart/, with a file inside called orca-autostart.desktop. Since Orca is the screen-reader program that GNOME uses, i figured this file was there so that the display manager could be read out to people who couldn’t read the screen. However, i wanted to know if i could throw other .desktop files in there to make them start automatically as well. This lined up with another article i saw about playing a sound when you logged in on Ubuntu.

Since i didn’t have this “Startup Applications” program sitting around to configure it visually, i would have to create this entry myself and throw it in the right place. .desktop files are the Linux world’s answer to desktop shortcuts or application menu entries. It’s easy enough to make one of your own if you know the basic format and what you want it to run. Since the goal of this shortcut was “play a sound file”, i figured the thing to run would be something like paplay /path/to/sound-file.ogg. (paplay is a command to play an audio file with PulseAudio, a common audio system on Linux. It’s also the command used in the earlier post about Ubuntu.)

The file i came up with wound up looking like this:

[Desktop Entry]
Type=Application
TryExec=paplay
Exec=paplay /usr/share/gdm/startup.ogg
Terminal=false

Name=Play sound at startup

After writing this file to /usr/share/gdm/greeter/autostart/startup sound.desktop and copying the audio file to /usr/share/gdm/startup.ogg, i rebooted my computer to see if it would work. And… it did! Kinda!

The first version of the video in the tweet had audio that was a little thin-sounding compared to the final version. I didn’t realize it at first, and just thought that it was because the volume was low. After a while, i realized what was going on. It was playing through the speaker in my monitor! I didn’t want this, primarily because it sounded worse than going through my regular speakers, but also because it meant i couldn’t mute it by turning off my speakers, like i would regularly. So i had to figure out what was going on.

I’d seen a similar situation on the common-use Ubuntu computer i mentioned, where it would default to an unused audio device instead of the built-in speakers, so i tried looking at those avenues first. PulseAudio has a “default preferences” file it loads, so i tried to set the default audio-output device there, by adding a set-default-sink <analog stereo line-out device> line to /etc/pulse/default.pa. Unfortunately, this didn’t work, much like it didn’t on the Ubuntu computer.1 I also tried disabling the HDMI output device by settings its profile to off, but that only caused it to mute the sound when it was trying to play.

Eventually, what i had to do was to sign in as the GDM user using the command listed in the Arch Wiki page (machinectl shell gdm@ /bin/bash) and try to configure PulseAudio from there. Unfortunately, this hit another hurdle because PulseAudio is only allowed to be running from one user at a time, and once i’ve logged in via GDM i’ve fully taken it over.

Instead i had to do this while GDM was still running, by pressing Ctrl+Alt+F2 to open a different TTY, and sign in there in a console session, so i wouldn’t try to start PulseAudio. From there i could open the shell with the GDM user as before and try to set things up. After confirming the setup with pacmd list-sinks and pacmd list-cards to look at what GDM was seeing from PulseAudio, i was able to disable the HDMI output with pacmd set-card-profile 0 off to turn off the HDMI output entirely, for the GDM user. After doing this, i rebooted again to see what that did.

And after that, it was still silent! But the problem then was that the volume slider in GDM itself had been dialed all the way down. After setting the volume from the system menu in the corner before logging in and rebooting once more, i was able to hear the beautiful sounds of the Windows 95 startup sound coming from my Linux computer in 2020.


I don’t have a moral here about things being complicated or open or whatever. I knew going in that this was a stupid idea, and i’m thankful that i was even able to pull it off. If i had tried this a few years ago, i would probably have given up midway through, because i wouldn’t have known to bother with the per-user PulseAudio settings or have wanted to bother with it. This wasn’t something that had a guide on the internet that had all the steps for my specific situation - for one, i’m running Arch Linux, whose entire ethos is some variation of “do it yourself”. That’s partly why i wanted to write all this down, in case someone else wanted to try the same thing.


  1. So what happened here was interesting, and leads into the eventual solution. Apparently, PulseAudio saves which audio device a given user used last in their user folder as the “default device”. So if you look in ~/.config/pulse/, you can see a couple files called <hash>-default-sink and <hash>-default-source. These files have the full names of the device that PulseAudio will try to load when that user starts up the PulseAudio daemon. The problem i had run into was, once these were set, just saying “the default device is this, actually” in the global config file wasn’t going to override that! That was part of why i needed to fully disable the HDMI output from the GDM user - it was still just going to use that device no matter what.