We run a simultaneous translation streaming system for a client who does world-wide live events that get streamed to many thousands of folks live around the world using vMix.
As we can be streaming up to 6 languages we use 2 PC's running 3 streams each. With Vmix 28 there is now scope to stream more outputs, but here's the problem.

Vmix doesn't do great audio features, we need to stream original floor video and audio with a translators voice on top of the original floor audio, we do this by ducking the floor audio. The ducking is controlled using the translators voice to trigger the compressor side chain that ducks the floor audio. You can't do ducking via a side chain in vmix.

To get around this problem I need to send all the audio to a DAW on another PC via a dante network, then bring the mixed audio back to vmix, sync it up then broacast it via amazon.

This of course uses up all the busses, 3 per PC to get the audio to the DAW, then 3 to do the broadcasting. We could do many of these events with 1 PC if there were more busses, 4 more would be ideal, then when we just do 4 languages we only need 1 PC. I wonder how the folks in the vmix R&D department might respond to this.

A second question, Were moving into a new era of technological posibilities with AI.

I am investigating a new translation system, we currently use an online system that is relativley poor quality, the new AI system would give us a cheaper and better quality output. To make this new system work it would be very beneficial to be able to do 1 transmission to a cloud service with audio only?

This raises 2 questions, one is how easy would it be to add a specific protocol that suits the AI system I'm looking at? The other is as I mentioed earlier, can it be audio only?
I better stop here, I can see a lot of questions coming my way asking what the sam hell am I talking about, I invite all the questions if we can reach a conclusion.

Many thanks
George.