Real time transcription | Twilio Agent Conference

I am referrring to the demo Kris gives here, especially the separate active transcriptions that show up for Stuart and Kris. I am familiar with conference, gather, dial but I am not able to replicate the entire architecture. I am interested in placing 2 people on a conference and somehow receive their transcriptions in real-time as the wonderful demo shows. I have enabled Agent-Conference tried the coach feature, but still can't figure out how to use Gather with conference participants. If I put a participant in Gather mode the other participant can't hear her. Can anyone please throw some pointers, I am sure I'm missing something basic here.

Till now I have tried:

Call call1 = Call.creator(new PhoneNumber("MyNumber"), new PhoneNumber("VerifiedTwilioNumber"),
                new URI("https://mywebsite/gather.xml")).create();

https://mywebsite/gather.xml returns

<?xml version="1.0" encoding="UTF-8"?>
<Response>
<Gather input="speech" speechTimeout="auto" partialResultCallback="https://mywebsite/gather_twilio_webhook_servlet">
    </Gather>
</Response>

Using this I speak to no-one but whatever I speak gets shown real-time on my gather_twilio_webhook_servlet.

I have also made conferences using :

Call call1 = Call.creator(new PhoneNumber("MyNumber"), new PhoneNumber("VerifiedTwilioNumber"),
                        new URI("https://mywebsite/conf.xml")).create();
Call call2 = Call.creator(new PhoneNumber("MyFriendsNumber"), new PhoneNumber("VerifiedTwilioNumber"),
                        new URI("https://mywebsite/conf.xml")).create();

https://mywebsite/conf.xml returns:

<?xml version="1.0" encoding="UTF-8"?>
<Response>
  <Dial>
    <Conference startConferenceOnEnter="true" endConferenceOnExit="true">Room 123</Conference>
  </Dial>
</Response>

I am trying combinations thereof but not able to find a documentation to replicate Kris's exact use case. Any help will be much appreciated. Thanks!

Update, more trials:

If I try something like:

<?xml version="1.0" encoding="UTF-8"?>
<Response>
    <Gather input="speech" speechTimeout="auto" partialResultCallback="https://mywebsite/gather_twilio_webhook_servlet" language="en-IN">
        <Dial>
            <Conference record="record-from-start">room-2</Conference>
        </Dial>
    </Gather>
</Response>

I get ErrorCode "13320" Msg "Gather: Invalid nested verb"

Twilio developer evangelist here.

The demo Kris showed is not available right now to use publicly. I know it was a while back, but bringing it to market was not as high a priority as other things we have launched since.

If you are interested in features like this, I recommend you get in touch with sales and discuss what you would like to use it for.