Friday, 17 August 2012

Hearing Voices - part three

I've now dug out my USB headset and configured it on the flightsim PC.  I have the FS sound effects on the PC main speakers, the ATC comms traffic through the headphones and the co-pilot voices (if any) through the secondary speakers.

Because the Multi-Crew Experience demo is the shortest (four days or thirty flights) I tried that one first.  Before doing anything I reran the default MS speech recognition training four times (twice each through the "tips" and "background" sessions), then the VoxATC training, and all four of the Multi-Crew Experience training sessions.  It's worth noting that the MCE training is quite extensive, including a large number of "real" ATC responses with various callsigns and situations.  VoxATC offers a general training session, and also allows you to setup a session based on a given flightplan, which I assume would include destination and waypoints, handy if you're an Aussie simmer flying out of Wollabrawonga aerodrome, or somewhere similarly unpronounceable.

Using the USB headset did pretty much eliminate the problem with the copilot overhearing ATC chatter as commands.  But sadly the reliability did still leave a lot to be desired.  Here's a real transcript of one part of the test flight I did, just after takeoff, gear up and climbing through 2000.

Me: Set altitude to eight thousand
Trav: Roger, setting altitude to twenty-eight thousand.
Me: (sigh) No, set altitude eight thousand.
Trav: Roger altitude set to eight hundred.
Me (as the plane takes a nosedive towards the South Manchester suburbs): Noooooooooooo

 I'd say the recognition rate was no better than 50%.  Bear in mind that this was after quite a bit of voice training.  Some functions worked better than others ."Set Com One to XXX decimal YY" worked correctly about 80% of the time.  But requesting checklists only worked about 10%, and unlike It's Your Plane, MCE's copilot isn't proactive and won't suggest actions unless you initiate them.

Another problem I had was that Trav seemed to go daydreaming for some periods.  I'd ask for something and get no response, then after two or three requests I'd get several answers from him at once, most of them variations on "Huh, what did you say?", sometimes interspersed with a snarky comment.

Trav's personality was also becoming a bit of a problem.  The MCE programmers have done a wonderful job in making him seem human, and several times I found myself talking back to him for fun like I would banter with a real-life work colleague.  When he makes a justifiable snarky comment about your flying, it's rather fun and useful - for example he pointed out that my 40kt taxi-ing was a little too fast for comfort.  But he quite often seems to mishear things and interpret them as profanity, which gets a little tiresome.  And worse he sometimes decides to chip in with some "human" comment right when things are tting a little busy and you really need him to SHUT THE HELL UP AND LISTEN!  Seriously, the developers need to add a "professionalism" tickbox in the control panel to turn off the funnies for when they get old. (edit - I've just seen in an online forum that there is an option to tell Trav to "stop nagging" or disable it permanently in an .ini file.)

ATC still suffered from the problem of crashed calls.  It was soul destroying to read back a detailed response, have the option line flash to show that it had been recognised, then have an AI voice come on frequency before it could be transmitted.  Again, recognition rate was not great in the first place and in the end I started giving each response two tries before hitting the key to manually set the option.

I've got today and Saturday left on the MCE trial, plus a couple of days after that on the VOXATC demo.  Trying MCE with just the ATC enabled (no snarky Trav) is next on the aggenda, followed by a VOXATC run using the headset mike.  The Its Your Plane demo seems to be time unlimited so I'll give that one more try as well, but since I can't test the SuperATC function I'm leaning away from that choice.

Impressions so far: VoxATC seems to be the most reliable as far as ATC voice control went.  Even though it was just some simple test flights and I was mainly just reading the scripted responses, at least I was able to get up into the air and follow directions successfully with it.  The more realistic comms procedures aren't a bonus for me.  To be honest the procedures in the default MS ATC are quite enough to provide verisimilitude for me.  The big downside is the cost - VoxATC is expensive to start with, and really needs a number of add-on voices at £25-30 a pop to make it feel like there isn't an army of cloned Annas running the MSFS world's ATC network.

MCE offers an option for the default ATC, but with problems (see above).  It does however come with a base set of high quality recorded voices AND has the option of adding voice fonts as alternatives.  Not only that but the developes are promising to add a number of additional voices in the coming months as free updates.

Both do the core thing I want (i.e. speaking to ATC) - VoxATC is more scripted and uses "real" procedures, MCE is more open in the choice of words and uses the default MS ATC.   Vox ATC has the advantage in avoiding "crashed" comms (where someone else comes on frequency blocking your transmission)

Both also do something in addition to the core of what I want.  VoxATC does AI traffic generation, which I already have with TrafficX, but it does it well.  MCE adds a copilot function, which would be useful, but so far seems more trouble than he's worth.

Interestingly enough, I'm finding the hassle of trying to get the speech recognition to work rather detrimental to the sense of immersion.  Trying two or three times to say a response then hitting the option key in frustration does more to take me out of the simulation than simply hitting the button and hearing "my" voice read it out would.  Maybe that would get better with practice and more voice training, but for now I'm seriously questioning whether the hassle is worth it.

It's funny, the quest for "as real as it gets" means so many different things to different people.  I know some simmers take real-world ATC audio from LiveATC and play it while simming.  For me, listening passively to ATC chatter that I can't interact with and has nothing to do with what's going on in the simulated world would feel less realistic, but for them I guess it's more.  For other people it's all about raw frame rates, and they talk about making it more realistic by turning graphic details down or off.  And then there are those running incredibly detailed aircraft sims which emulate every last detail of an aircraft's operation.  Personally I'm less interested in sitting at my PC for 10 minutes while the gyros align themselves, or have to remember to turn on power to the cabin galley.  I want just enough detail to let me feel I'm running a complex aircraft, without requiring more than a couple of hours reading manuals and running through tutorials.

Your air-miles may vary.


No comments:

Post a Comment