This is probably premature, but I was curious to see how TTS has evolved since I last researched it several years ago. Lo and behold, Azure seems to have taken things up a notch or two using neural networks. Even "standard" TTS tech seems to have made great strides in believability. 

Take a look here: 

https://azure.microsoft.com/en-us/services/cognitive-services/text-to-speech/

Do you think MS will leverage Azure TTS for ATC in the new MSFS? I think it would greatly reduce the amount of effort required to implement ATC--no need to hire talent, book studio time, and record thousands of words and phrases per voice--thus freeing up development resources and making it more likely that we see some of the more advanced ATC features users have been pining for. It would also be much easier to expand ATC functionality over time, or amend existing procedures and phraseology.

What about TTS for other uses? Perhaps it could be used in the GUI, or during avatar interactions outside the cockpit? Perhaps it could be a valuable tool for mission designers to create dialogue, briefings, etc?

I'm curious to know if other users want to see this implemented or not?

Personally, I like my FS experience to be a single-player one. I thought FSX ATC was surprisingly versatile and convincing--especially after using EditVoicePack--and I'd like to see that system improved and modernized. Although multiplayer services like VATSIM and PilotEdge are very cool, they all share the inherent problem of a lack of coverage/availability. I want to fly any time of day or night, to any city or place in the world, and know for certain that Clearance Delivery and Ground/Local, and Approach/Departure and Center/Flight Watch are all manned...something that only integrated single-player ATC can accomplish.

I always thought in favor of human-like voices over realistic procedures. It's great to get clearance to descend via STAR X but if that voice is totally robotic it kills all the immersion for me. Hence I was never a fan of TTS voices. The sample sentences on the top of the page were pretty good, though. They sound very fluent without the 'hiccups'. I'm nonetheless a tiny bit skeptical about the samples because it's just three sentences. I want to know what the voices sound like when it's putting together a long string of instructions, e.g. when obtaining clearance in the sim. Also, for a realistic experience the voices would have to sound like they are actually coming over radio. Not as exaggerated as in some addons which sound like aircraft are still using 1940s radios, but the distinctive sound of ATC transmissions would have definitely have to be there as well.

I have been using or I might say enjoying flying with flight sims since 1995 and you all guessed it Microsoft all the way from the start I used the ATC that was provided and today I use pilot 2 atc since x plane 11 past x plane 9 their atc is not that good at all so pilot2 atc is what is going on for now.

You can’t delete posts on here?

Edited August 11, 20196 yr by Superdelphinus

This could be a true revolution! This is what we always wanted.

Hope they will integrate this technology in the sim!

23 hours ago, threegreen said:

I always thought in favor of human-like voices over realistic procedures. It's great to get clearance to descend via STAR X but if that voice is totally robotic it kills all the immersion for me. Hence I was never a fan of TTS voices. The sample sentences on the top of the page were pretty good, though. They sound very fluent without the 'hiccups'. I'm nonetheless a tiny bit skeptical about the samples because it's just three sentences. I want to know what the voices sound like when it's putting together a long string of instructions, e.g. when obtaining clearance in the sim. Also, for a realistic experience the voices would have to sound like they are actually coming over radio. Not as exaggerated as in some addons which sound like aircraft are still using 1940s radios, but the distinctive sound of ATC transmissions would have definitely have to be there as well.

Having an audio distortion layer to mimc ATC interference play over the top of a TTS track is pretty straight forward. I don't think the ATC anouncer would need to change there voice to accomodate interference, infact I suspect they are encouraged to speak as naturally and camly as possible, exactly the tone and pace of the TTS examples, seems like a good solution.

Where it gets interesting would be when you start doing things which could anoy the ATC controller and you really want them to just go nuts at you... 😁 but hey maybe thats simulated as well...

3 minutes ago, dtrjones said:

Having an audio distortion layer to mimc ATC interference play over the top of a TTS track is pretty straight forward. I don't think the ATC anouncer would need to change there voice to accomodate interference, infact I suspect they are encouraged to speak as naturally and camly as possible, exactly the tone and pace of the TTS examples, seems like a good solution.

One thing that would add realism is to slightly trim the beginning and end of each voice playback, to simulate the way people are often a little late hitting the PTT switch when starting to talk, and will often hit it a little early when they stop talking. It's how you get that "clipped" effect when hearing real ATC conversations.

That might not be ideal for newbies learning ATC coms, but if the processing is done in real time by AI, it should be possible to offer an option for slower full speech,or a more realistic-sounding rapid and truncated synthesized ATC voice.

I watched a video online the other day that suggested that the future of ATC probably wont involve voice interaction at all. Instructions and acknowledgements are sent and received via a text system.

5 minutes ago, sightseer said:

I watched a video online the other day that suggested that the future of ATC probably wont involve voice interaction at all. Instructions and acknowledgements are sent and received via a text system.

Maybe in the future, but this is a simulator and today ATC is made via voice so it has to be as close as possible to the real world.

Edited August 11, 20196 yr by Noooch

I think the real power in using this service is the ability to apply regional variations using the same ATC text. This page appears to suggest regions are already a thing and it would be.. well incredible if something like this could find it's way into a Flight Simulator. I know this was something Stephen Hood was very excited to deliver to FSW.

https://docs.microsoft.com/en-us/azure/cognitive-services/speech-service/regions

Speech-to-text, text-to-speech, and translation

The Speech SDK is available in these regions for speech recognition, text-to-speech, and translation:

Region	Speech SDK Parameter	Speech Customization Portal
West US	westus	https://westus.cris.ai
West US 2	westus2	https://westus2.cris.ai
East US	eastus	https://eastus.cris.ai
East US 2	eastus2	https://eastus2.cris.ai
Central US	centralus	https://centralus.cris.ai
North Central US	northcentralus	https://northcentralus.cris.ai
South Central US	southcentralus	https://southcentralus.cris.ai
Central India	centralindia	https://centralindia.cris.ai
East Asia	eastasia	https://eastasia.cris.ai
Southeast Asia	southeastasia	https://southeastasia.cris.ai
Japan East	japaneast	https://japaneast.cris.ai
Korea Central	koreacentral	https://koreacentral.cris.ai
Australia East	australiaeast	https://australiaeast.cris.ai
Canada Central	canadacentral	https://canadacentral.cris.ai
North Europe	northeurope	https://northeurope.cris.ai
West Europe	westeurope	https://westeurope.cris.ai
UK South	uksouth	https://uksouth.cris.ai
France Central	francecentral	https://francecentral.cris.ai

16 minutes ago, dtrjones said:

I think the real power in using this service is the ability to apply regional variations using the same ATC text. This page appears to suggest regions are already a thing and it would be.. well incredible if something like this could find it's way into a Flight Simulator. I know this was something Stephen Hood was very excited to deliver to FSW.

https://docs.microsoft.com/en-us/azure/cognitive-services/speech-service/regions

Speech-to-text, text-to-speech, and translation

The Speech SDK is available in these regions for speech recognition, text-to-speech, and translation:

Region	Speech SDK Parameter	Speech Customization Portal
West US	westus	https://westus.cris.ai
West US 2	westus2	https://westus2.cris.ai
East US	eastus	https://eastus.cris.ai
East US 2	eastus2	https://eastus2.cris.ai
Central US	centralus	https://centralus.cris.ai
North Central US	northcentralus	https://northcentralus.cris.ai
South Central US	southcentralus	https://southcentralus.cris.ai
Central India	centralindia	https://centralindia.cris.ai
East Asia	eastasia	https://eastasia.cris.ai
Southeast Asia	southeastasia	https://southeastasia.cris.ai
Japan East	japaneast	https://japaneast.cris.ai
Korea Central	koreacentral	https://koreacentral.cris.ai
Australia East	australiaeast	https://australiaeast.cris.ai
Canada Central	canadacentral	https://canadacentral.cris.ai
North Europe	northeurope	https://northeurope.cris.ai
West Europe	westeurope	https://westeurope.cris.ai
UK South	uksouth	https://uksouth.cris.ai
France Central	francecentral	https://francecentral.cris.ai

Does this mean different accents in english or different languages?

20 minutes ago, Noooch said:

... and today ATC is made via voice ...

Not correct. Increasing amount of communication is now provided through CPDLC (Controller-Pilot-Data Link Communications), even in high-density European airspace.

https://ext.eurocontrol.int/WikiLink/index.php/Current_Implementation_Status

36 minutes ago, SierraDelta said:

Not correct. Increasing amount of communication is now provided through CPDLC (Controller-Pilot-Data Link Communications), even in high-density European airspace.

https://ext.eurocontrol.int/WikiLink/index.php/Current_Implementation_Status

Those are only centers, what do you do with towers, ground, arrivals...?

7 hours ago, Superdelphinus said:

You can’t delete posts on here?

You can't but I can. What is it you need?

 

The use of ADS-C and CPDLC is definitely expanding, especially for trans-oceanic flights, but the vast majority of control takes place via VHF Voice.

The regional accents are pretty convincing, and it would be great to have French-accented controllers in France, and Spanish-accented controllers in Mexico, etc... Maybe even slip in the likelihood that controllers will occasionally speak the local language when talking to local airlines. For example Paris Ground speaking to a Air France flight in French. 

I think adding a radio filter to the voices in real-time is doable. Perhaps even a variable effect based on factors like line-of-sight, distance, and atmospheric effects. Might be a candidate for a multithreaded task.

Given the entirely synthetic nature and adaptability of TTS, it might be possible to do some cool stuff like increase the speed of controller speech at certain facilities, or even do so dynamically based on traffic load.

Something I'd like to see is occasional simultaneous communications and the resulting heterodyning interference. Just little things that add ambiance and realism go a long way toward suspending disbelief.

 

Edited August 11, 20196 yr by Noodle