Wednesday, March 14, 2018

Simultaneous Translation Tech Pilot

When we moved to Cluj, we started attending a local church, Casa Tamplarului.  As a church, it reminds us a lot of National Community Church in DC, which we attended before we moved here.  Of course, being the audio nerd that I am, I’ve joined the production team and now run front of house regularly.

As much as I like the church and the people there, I have one issue: the message is delivered in Romanian.  Actually, it’s not a problem with the church, but it’s really my problem as my Romanian isn’t quite as far along as I would like.  I’m learning, but I’m not there yet.

In talking with some of our friends there, the team has wanted to offer simultaneous translation of the service into English for a while.  They even started collecting some consumer-grade wireless headphones to try and use.  While we were able to make them function, they offer some limitations, so we started looking at alternatives.

What we’ve settled on as an initial trial is to use mobile phones and streaming audio as a delivery platform.  It offers a few key advantages:

  • Everyone has a mobile phone, so we are not limited to hardware that the church has to provide.
  • Phones are already set up to receive streaming audio, the listener just needs to download an application.
  • As we move forward, it offers the option of broadcasting the audio not only locally but to the Internet at large as well.
  • If we decide to add more languages, it should be just a matter of adding additional streaming endpoints to the existing tech.

For the initial pilot, we started with a completely local setup with the following components:

  • One Windows laptop
  • One Mac
  • Google WiFi access point(s)
  • An Icecast server on the Windows laptop that would offer up audio streams to as many endpoints as possible.
  • Ladiocast on the Mac which would forward the audio to the Icecast server.
  • TuneIn (Android, iOS) as the streaming client on each of the mobile devices.

Network topology

One key thing I did was to wire up everything that was static.  Both laptops were wired into a gigabit desktop switch in order to limit the amount of traffic on the wireless network.  We also did not connect the access point to the Internet, mostly because it was an initial test, however, as I’m thinking about it now, we may not want to connect it to the internet at all, again in order to limit saturating the wireless network.

In the end, with just me connecting, the technology did work.  There is a fairly long delay of about 10-15 seconds in the audio stream.  I think this is due to the client buffering the audio.  I will continue to look into how to reduce that latency, but for now it’s workable.

2018 03 11 11 51 00 1

We plan on trying it again this weekend with a larger audience of clients and see how it goes.  We also need to gather together some additional upstream tech (microphones, mixers, etc) in order to complete the rig.

I also plan on updating things here as we get closer to a finished solution and start learning lessons of what to do and what not to do.

No comments: