Vocal Isolation

[Garbaz]

I'm looking for a way to achieve an effect like this one in Audacity on the fly, i.e. to isolate speaking voices from things like background music. The effect is relatively fast to run, so I would imagine it's not something overly complex that can't run in real-time.

Is this possible with the current toolset available in Easyeffects, at least to some degree? If not, is there some manual way to achieve this? If not, consider this a feature request.

[wwmm]

Maybe if the noise reduction plugin is used in combination with the Stereo Tools plugin something similar can be obtained in some situations if the signal to noise ratio leads the noise reduction plugin to consider the background music as noise.

[Garbaz]

I've tried your suggestion, but I couldn't get the Noise Reduction plugin to behave, it always produces crackly or splotchy sound, though I'm not sure in the first place how I could suggest to it that the background music is noise, at least not in any consistent manner.

There is this LV2 plugin, which seems to do a similar thing to Audacity. Could this be added to Easyeffects? I've tried it in Audacity, and it seems to do the job.

[Digitalone1]

As a workaround, you could load that plugin in carla rack under jack-pw and use it through pipewire.

Here is an how-to.

[wwmm]

There is this LV2 plugin, which seems to do a similar thing to Audacity. Could this be added to Easyeffects?

We are able to use most of the LV2 plugins. It depends on what the plugin requires. I will have to do some tests.

[Garbaz]

Alternatively to kn0ck0ut, the audacity effect is here. It's a "nyquist plug-in" though.

[Garbaz]

As a workaround, you could load that plugin in carla rack under jack-pw and use it through pipewire.

Here is an how-to.

After some head-scratching, I got it to work. And it works great! Thank you very much.

For anybody else trying to do this:

Note: You have to change in the settings of Carla in Engine the Audio Driver to JACK. For me it was by default set to Pulseaudio.

My script for starting it:

if [ -z "`pactl list short  | grep 'default_null_sink'`" ];then
    pactl load-module module-null-sink object.linger=1 media.class=Audio/Sink sink_name=default_null_sink channel_map=FL,FR
fi
pw-jack carla-rack

[wwmm]

Our master branch now have a Voice suppressor plugin. I've added an Inverted mode to it so we can do the opposite operation. It is not perfect. But voice isolation without machine learning is tough. But the plugin should still be useful in simple cases.