https://xyproblem.info/

Don't give people your solution and ask them how to do it. Start with your problem out of the gate.

Instead of checking for audio maybe you can write a usescript to run actions based on what's happening on the website. Dunno tho cuz Im making assumptions at what the problem is.

This smells of "So I work from home, but want to sleep, but if my boss pings me on teams I want an alarm to wake me up"

Be careful: this could be you https://xkcd.com/1172/

XY problem?

I think it would possibly be easier to write an extension. You can inspire yourself from this extension, intercept media playback, possibly hash the media being played, compare the hash to a known DB you create, and call a script in response to a positive detection.

CC BY-NC-SA 4.0

Someone already talked about the XY problem, so I'll say this.

Why sound notification instead of notification content? If your notification program (dunst in my case) have pattern matching or calling scripts based on patterns and the script has access to which app, notification title, contents etc. then it's just about calling something in your bash script.

And any time you wanna add that functionality to something else, add one more line with a different pattern or add a condition in your script. Comparing text is lot more reliable than audio.

Of course your use case could be completely different, so maybe give some examples of use case so people can give you different ways to solve that instead of just the one you're thinking of.

I have some vague recollection of a hacker convention from the 90s where people were challenged to come up with wireless networking in a one night coding marathon. (This was long before wifi.) So some dude used speech synthesis to get a machine to say "one zero one one zero…" and another to assemble the binary data into packets using speech recognition. It was hilarious, and the dev had to keep telling people to shut up and stop laughing so he could complete the demo.

But anyways… what I'm trying to suggest here is you might have the best luck if your notification sounds contain spoken commands and you use speech recognition to trigger scripts? That tech is pretty mature at this point.

Maybe you can do something with the tampermonkey extension to catch when that audio is triggered and have it do an api call that your script catches?

I don't know if that'll actually work, I know of the extension but have never it used nor am I skilled with Javascript but it seems feasible.

Is the sound screaming?

Can you isolate the call to the sound from the DevTools? And if so, does DevTools allow you to edit the function? Perhaps you could GET/POST something on localhost which could trigger a shell script.

I think sox is what you want

I went ahead and asked a free AI how to use sox and played audio to trigger a script. Probably won't be 100% accurate but maybe send you down the right path. Good luck.

1. Install 'sox' if you haven't already. You can use the package manager of your Linux distribution to install it.

2. Open a terminal and use the 'rec' command from 'sox' to continuously listen to the audio input:

   sox -d -t .wav - silence 1 0.1 3% 1 1.0 3%
   

   This command will listen for audio and create a .wav file when it detects sound.

3. Write a script that will be triggered when a sound is detected. For example, you can create a script called "myscript.sh" with the following content:

   #!/bin/bash
   echo "Sound detected! Running my script."
   # Add your desired actions here
   

4. Make the script executable by running the following command in the terminal:

   chmod +x myscript.sh
   

5. Use the 'rec' command along with the 'play' command from 'sox' to continuously listen for sound and execute your script when sound is detected:

   sox -d -t .wav - silence 1 0.1 3% 1 1.0 3% | while read -r; do ./myscript.sh; done
   

   This command will continuously listen for sound and execute your script each time sound is detected.

Remember to customize the script "myscript.sh" with your desired actions.

Expect automates things, based on text-input captured from a terminal.

Not sure if it has been extended/hacked to take sound as an input.

My assumption is that you don’t care if your notification gets spoofed, ex. Someone rings a little bell and the script deletes all cookies from porn websites as if the little bell notification played.

So I think the hardest and best way to do this is to have the script run on a separate device than the sound plays on.

First record the sound you want to trigger with. Use the script executing device with the microphone and interface you’ll be using in production set up in the location of production to make it easier on yourself.

Now reduce the bitrate of the target sound a lot. No, more than that, keep going, a little more, that’s perfect.

Now write something that will capture the last target_sound_length seconds of audio and compare it with the bitcrushed version. Depending on your device, there may be a buffer object in the adc you can interface with, although if it’s running a normal operating system you won’t be able to just get to it without going through the os first.

If you can go through the chrome notificationing machine, figure out the hook used to trigger the notification you want to respond to and intercept and perform the script. No nyquist needed!

Sounds like a case where machine learning would actually be useful.