Add "audio-only" input option

[crossan007]

Currently, the plugin supports audio as part of the source, but an audio-only plugin would be nice so that when the scene no longer includes the source, the pipeline may still render audio into the mixer.

Specifically, this pipeline comes to mind:

videotestsrc pattern=18 ! video. alsasrc name=audiosrc device=hw:1 ! audio/x-raw,channels=32,layout=interleaved,rate=48000 ! queue ! deinterleave name=d interleave name=i d.src_0 ! queue ! audioconvert  ! i.sink_0  d.src_1 ! queue ! audioconvert  ! i.sink_1 i. ! audio.

I'd like to not need to map anything to the .video sink, and allow this pipeline to run regardless of the currently live scene (and whether the obs-gstreamer source is present on that scene)

[fzwoch]

Isn't that already working? Having an audio only pipeline? At least that was my intention. Remove any video paths and the sound should still work. Or did I not understand your use case?

[fzwoch]

Ah I think this request for acting as an input device not a source. I need to check how that works inside OBS though..

[fzwoch]

Does such a thing exist in OBS studio? Looking around I figured - all plugins are global. I don't think I can add a "driver" to plug in into the desktop audio devices etc. The plugin code mentions audio in combination with video most of the time.. can elaborate what exactly you had in mind?

[crossan007]

My use case: 32 channels coming in on USB from an audio mixer.

I want to prune off two of the channels to be used as a constant OBS audio source.

I could select the device from the normal OBS audio input device selector; however, that doesn't give me any control over which of the 32 channels are used in the 2 channel output

[fzwoch]

This sounds like a limitation on the OBS input selection then?

Alternatively you use this one but you would need to add it to every scene. But I don't see a way to "bridge" these two..

[crossan007]

yeah - it would be nice if OBS allowed configurable mapping of audio inputs, or a plugin architecture for persistent audio inputs.

My current setup has this GST pipeline defined once in a scene, and then I include that scene in every scene that could go live.

There are two drawbacks with the "include everywhere" method:

• If a scene includes two other scenes which have the GST pipeline included, there is a net audio volume gain of a few decibels.
• If the GST pipeline was forgotten on a scene, then that will be a silent scene.

[fzwoch]

If a scene includes two other scenes which have the GST pipeline included, there is a net audio volume gain of a few decibels.

I guess that's because the same audio is added twice and accumulates to some higher level audio?

f the GST pipeline was forgotten on a scene, then that will be a silent scene.

Not sure if I understand that one.

Perhaps the real solution would be to output the audio via gst-launch-1.0 to the regular sound device (or a virtual added one?) and let OBS pick that up via the default sound device settings?

[crossan007]

I guess that's because the same audio is added twice and accumulates to some higher level audio?

I think that's what's going on.

If the GST pipeline was forgotten on a scene, then that will be a silent scene.

You might be over-thinking this one ;-) If I forgot to add the audio pipeline to a scene, then the audio obviously won't be in the mix, so the scene is silent.

Perhaps the real solution would be to output the audio via gst-launch-1.0 to the regular sound device (or a virtual added one?) and let OBS pick that up via the default sound device settings?

Yeah, this is probably better.

[crossan007]

I've been looking through the OBS code on this issue, and made an interesting observation.

It looks like the "bundled" audio input sources (available from the OBS Settings | Audio menu are nothing more than standard OBS input plugins; however, the plugin available for use is hard-coded by a compile-time define statement: https://github.com/obsproject/obs-studio/blob/master/UI/obs-app.cpp#L1305

The name of the "selected" audio input plugin is serialized out to the DefaultCollection.json file in the OBS user directory by (https://github.com/obsproject/obs-studio/blob/master/UI/window-basic-main.cpp#L2238), and of specific interest is this: https://github.com/obsproject/obs-studio/blob/master/UI/window-basic-settings.cpp#L3270

I was able to copy/paste the JSON nodes for id as well as settings in my DefaultCollection.json file so that gstreamer-source may be used as an independent audio input without needing to add it as a sub-scene to all scenes that I want to use.

It's a bit hacky, but means that it may be relatively easy to extend the OBS window-basic-settings.cpp dialog to allow for additional "global" audio input plugins

With this, my DefaultCollection.json contains this:

{
    "AuxAudioDevice1": {
        "deinterlace_field_order": 0,
        "deinterlace_mode": 0,
        "enabled": true,
        "flags": 0,
        "hotkeys": {
            "libobs.mute": [],
            "libobs.push-to-mute": [],
            "libobs.push-to-talk": [],
            "libobs.unmute": []
        },
        "id": "gstreamer-source",
        "mixers": 255,
        "monitoring_type": 0,
        "muted": false,
        "name": "QU 32 MultiChannel Lines 31,32",
        "private_settings": {},
        "push-to-mute": false,
        "push-to-mute-delay": 0,
        "push-to-talk": false,
        "push-to-talk-delay": 0,
        "settings": {
            "pipeline": "videotestsrc pattern=18 ! video. alsasrc name=audiosrc device=hw:1 ! audio/x-raw,channels=32,layout=interleaved,rate=48000 ! queue ! deinterleave name=d interleave name=i d.src_30 ! queue ! audioconvert  ! i.sink_0  d.src_31 ! queue ! audioconvert  ! i.sink_1 i. ! audio. ",
            "restart_on_eos": true,
            "restart_on_error": true,
            "restart_on_update": true,
            "stop_on_hide": true,
            "sync_appsinks": true,
            "use_timestamps": true
        },
        "sync": 0,
        "volume": 1.0
    },

By hand-editing the JSON, we run the risk of losing the changes anytime the audio settings are changed, since the dialog box cannot resolve the source name:

However, accessing the obs-gstreamer properties dialog box works as expected from the "gear" icon in the OBS audio mixer pane:

Also of note is that any video written to obs-gstreamer's video. app sink will be overlaid on top of the final image. A curious side-effect that, while probably not intended, could actually be useful to someone, somewhere:

[fzwoch]

Do you feel like moving this to the Wiki? :-)