Add support for CLI operation without opening the UI

[cipriancraciun]

It would be lovely to have some CLI arguments to just load / unload the Pulse plugin, without having to resort to the UI.

This is especially useful for those that want to create quick shortcut scripts to run for example with dmenu, rofi or on a key binding.

[lawl]

Not sure about this one, I have two questions:

• What would be the benefit over loading https://github.com/werman/noise-suppression-for-voice with a script directly?
• How would you select which input device to load NT for?

[cipriancraciun]

What would be the benefit over loading https://github.com/werman/noise-suppression-for-voice with a script directly?

Well the short answer is that NoiseTorch seems quite well focused and simple to use tool. (I'll take a look at that project.) Thus my only answer to this question is that NoiseTorch is the first tool I've used to for voice activation, thus I would prefer to keep using it. :)

How would you select which input device to load NT for?

In the end, if you want to keep things simple and not implement an entire CLI, how about this:

• in the UI you can add a checkbox which states something like "dry-run";
• when checked you can display a text area which presents an encoded "token" (say Base64) of the latest action;
• thus when one presses Load plugin, the plugin isn't actually loaded, but instead a token is generated;
• the same for Unload plugin; (for reload it makes no sense;)
• that "token" contains all the data that your tool requires to configure and load / unload the plugin; also it could be "opaque" for the user, thus you don't need to commit on a protocol here;

Now the user can just run noisetorch execute {token} or create a shortcut for that. (If it's a load command, and the plugin is already loaded, just unload it before re-loading it.)

It's not nice, but it allows one to "snapshot" a given configuration and load it quickly; say implementing a kind of "presets" outside NoiseTorch itself.

[lawl]

I'll have to decline that suggestion for the UI, this seems overly complicated, sorry. NoiseTorch also uses werman's project which I've linked, including some changes I have already upstreamed, so there's no difference there really. NoiseTorch just abstracts away the pulseaudio plumbing required to set it up.

Thinking about it, these would be the benefits over using werman's LADSPA wrapper directly:

• It takes care of workarounds such as RLIMIT_RTTIME required on some machines.
• It takes care of the exact pulseaudio config used instead of leaving that to the user, which is still evolving.
• It could still abstract away some pulseaudio stuff and make it simpler.

That said, I'm torn at the moment if this is enough to justify making a CLI interface vs. telling people to just do the pulseaudio plumbing themselves.

[cipriancraciun]

What I've described was just a "quick hack" UI. In the end my feature request was just for a way (any way) to simply obtain the same loading / unloading effects without the UI, from the CLI.

[lawl]

Yes, understood, and I have a rough idea how a CLI could look like. But I want to think about this a bit more.

[cipriancraciun]

As a side-note, having these shortcuts might allow one to easily work around the high CPU usage issue as reported in #41.

[stemid]

Hi so NoiseTorch uses the werman noise-suppression-for-voice plugin. But when I tried the werman plugin I could not get proper audio. Only distorted.

If I use NoiseTorch, which is apparently just a wrapper, I get proper audio.

So my question is, what does NoiseTorch do different from the pacmd commands werman recommends? I would also really like to use the plugin directly.

Part of the reason for this is that NoiseTorch requests sudo when you enable it, why? I pressed Cancel and it still worked. So I denied it sudo and it still worked. Using pulseaudio should not require sudo.

Edit: Ok I read the docs now on why it requires sudo. Apparently my system does allow this loading so I just hit cancel.

I'd really just like to see the PulseAudio config that NoiseTorch implements so I can do it myself. I don't like having unnecessary big programs around.

[lawl]

@stemid please don't spam random issues with questions and create an appropriate issue if you have a question. I will flag this as off topic and hide it.

That said, you can find the module loading code here. https://github.com/lawl/NoiseTorch/blob/master/module.go that should answer all your questions.

[lawl]

This has been implemented by @chance-nelson and is now merged and shipped in 0.7.0-beta.