[Suggestion] use of TTS for Betty so player addons become seamless

YorrickVander · Post by **YorrickVander** » Fri, 8. Sep 17, 21:49

Pretty much as the title says. I've become very used to HCS Voice Packs in Elite Dangerous in combination with Voice Attack. My favorite is the 'Verity' pack that uses the same voice actor as the in game computer speech and it becomes a totally natural extension. As this might not be feasible for X4 perhaps the use of a TTS engine for Betty and a voice players can buy/license for 3rd party use in Voice Attack, mods and other TTS compatible 3rd party tools? It would be very cool to set up a seamless voice interface with Betty

RAVEN.myst · Post by **RAVEN.myst** » Fri, 8. Sep 17, 23:34

I think this is an excellent idea. For starters, it circumvents the problem of voice clips becoming obsolete with changes down the line, or the lack of applicable clips for new content (such as would be introduced by modders, for example, as well as new content from ES themselves.) It would obviate inconsistencies such as when new content is added and either the original voice actor is unavailable, or recording settings don't match (for example, Vestibule of Creation in X3AP.) Additionally, having that hint of synth strikes me as appropriate to a shipboard COMPUTER voice

(which by definition would be synthesized, right?)

JoeVN09 · Post by **JoeVN09** » Sat, 9. Sep 17, 01:11

I like it. Save money on a voice actor and enable updates and mods without jarring incongruities in sound.

On the other hand, I'd miss X2 Betty's sexy voice.

pjknibbs · Post by **pjknibbs** » Sat, 9. Sep 17, 08:45

Have they managed to get text-to-speech engines that actually produce human-sounding speech yet, though? Even satnavs, which you'd think would be a shoo-in for the technology, usually use recorded voice samples, and when they don't, the mispronunciations can be a bit silly.

Artean · Post by **Artean** » Sat, 9. Sep 17, 09:06

There's only one Betty.

Morkonan · Post by **Morkonan** » Sat, 9. Sep 17, 20:09

TTS, really good TTS, is still a pretty high-end sort of thing. There's a good bit of research still being done. There are some open-source projects, though.

I wouldn't mind decent TTS, but I'm not sure the expense is justified. That's an Egosoft decision. If they got it wrong, the player would be irritated every time "Betty" said something. If they got it right, they might save some money for expansion pack production. So it's sort of a cost-benefit analysis based on expected future productions. While I think mod support should definitely be part of their focus, I dunno if the front-end cost justifies it. I wouldn't be mad if it just turned out to cost too much to get a professional quality TTS, so base-game players would be happy, just so mods could sound "more professionally polished."

PS - I, too, like Betty's voice and wouldn't mind her returning for X4, if possible. They could resample the originals and add the new ones, saving some costs I would think. Do some soundwork to get it all evened out, since even a voice-actor's voice/pronunciation can change a little bit, over the years. /shrug

Post by **UniTrader** » Sun, 10. Sep 17, 00:45

other suggestions/ideas:

=> when making the contracts for voicing maybe include an Option for additional voicing with the same VAs and settings/equipment some time after Release (eg maybe 2 years) and some time before that happens ask around in the community what additional texts should get voiced for the current Mods.

=> maybe combine the best of both really spoken Text and TTS: get VAs for Betty and other regulary used characters who also lent their Voice to TTS Engines, and use their natural Voice for the Spoken Text delivered with the Game, but also have the option of creating new snippets via TTS if needed and new recordings are not viable (and also publish or include the related information so there is no problem with adding new Samples)

RAVEN.myst · Post by **RAVEN.myst** » Sun, 10. Sep 17, 01:16

On a related note, but not specifically ship-computer: in Rebirth, one of the things I find most annoying (*in part* because it gets in the way of ship/other controls) are the long annoying monologues from NPCs when giving the most trivial responses. I would very much like to see an option in the audio settings for brief or verbose NPC dialog. Then, once a player has gotten tired of the equivalent of "How does your copilot cope with the smell of sweaty male clothing?" (as if clothing has gender, for one thing

) and has experienced the "flavourful" dialogue, he/she can then switch to an acknowledgement scheme that simply says "I'm aboard" or "I've arrived" or "I'm here" (hell - the brief acknowledgements can even have some variety while retaining brevity, as just demonstrated!), thus not wasting time while imparting the relevant info. Similarly, station managers reporting profits can stow those long-winded speeches and simply say something along the lines of "I'm sending you profits" or "Income generated and transferred" - and this same approach could be applied across the board, and perhaps even to some of Betty's announcements. In some specific cases, perhaps, the succinct voice option might altogether forgo some totally superfluous speech - for example, when turning off any scanner mode, BOTH Annoyisha AND Betty see fit to tell me about it, and rather languidly too - unnecessary! Of course, that example is specific to Rebirth, but analogous examples will probably show up in Foundations too.

caleb · Post by **caleb** » Sun, 10. Sep 17, 01:40

I love the idea!!! And if worried about quality, did some testing with a random text to speech site I found (not affiliated with them in any way, just the 1st one I found):

https://www.naturalreaders.com/online/

Here are some sample phrases I tried, and they sound pretty decent:

Your ships are under attack! in sector Argon prime!
Your base in cloud nine is out of resources.
Your ship, Nova Mark3 is out of ammo.
Boarding parties away!
Boarding parties have encountered heavy resistance!
Boarding parties have successfully captured the engineering deck!
Boarding parties are being pushed back, shield deck lost!
Boarding parties have successfully captured the bridge, the ship is ours!
Boarding parties have failed, remaining troopers are retreating!
Your ship Dolphin mark2 has sold 300 energy cells for 125326 credits.

So I would be 100% on board with this. And would help to have a computer that would actually give you useful data, and not just the same line over and over again...

gbjbaanb · Post by **gbjbaanb** » Sun, 10. Sep 17, 18:24

That TTS site was brilliant. UK Serena sounded like a newsreader. What's best about it is such a thing could give multiple voices, not just 1 from Betty (much as I love her voice).

RAVEN.myst · Post by **RAVEN.myst** » Sun, 10. Sep 17, 19:11

gbjbaanb wrote:What's best about it is such a thing could give multiple voices.

Which means that if the tech were to be embraced, it could be used to voice generic NPCs, too, thus creating the more diverse and meaningful speech that caleb mentioned. Then, for example, station profit transfers could be set to round numbers only, so you'd have "three hundred thousand credits deposited" or "five million credits in profit transferred" messages, and the NPCs could even identify themselves vocally ("Nyota Grey on Mining Base One here..." etc) - though here we run the risk of the transmissions once again becoming too lengthy (though if meaningful, that may make them less annoying, and worth the extra verbiage.)

SamuraiProgrammer · Post by **SamuraiProgrammer** » Tue, 12. Sep 17, 06:04

I think this would be awesome

Post by **UniTrader** » Tue, 12. Sep 17, 20:10

for generic voicing it might be ok-ish, but its not something which should be relied on entirely imo... for example i dont think TTS would work if an ironic under-tone is needed for some words in a sentence, Or when the speaking character is surprised by something mid-sentence, and stuff like that. Or it will require additional effort and more input to the TTS engine than just the words. then you can even let it sing.

RAVEN.myst · Post by **RAVEN.myst** » Tue, 12. Sep 17, 22:17

@UniTrader: I completely agree - hence "for generic NPCs", such as generated vendors, generic mission givers, etc. However, story NPCs should always be voice acted (unless, of course, it's an AI character that can benefit from sounding stilted, or perhaps an alien speaking through a translator - that would in fact add immersion, if the speech was patently synthetic-sounding.) And yes, even the above isn't all-encompassing - some "generic" NPCs might have to emote vocally, for example station security officers or what-not shouting at the player for doing unwelcome stuff in the vicinity, and the like...

Police could be mixed - after all, some (or perhaps even all) coppers could be robotic (it IS a dangerous job, after all...), and so a mechanical voice might not be out of place. In these cases, in fact, a flat deadpan expressionless monotone saying something that one would expect to come with excitement, could actually sound quite eerie and unnerving... "Come quietly or there will be... trouble."

birdtable · Post by **birdtable** » Tue, 12. Sep 17, 22:23

"Dead or alive ..Your coming with me"

https://www.youtube.com/watch?v=h4pIPXX2bYs

Rabiator der II. · Post by **Rabiator der II.** » Wed, 13. Sep 17, 17:37

pjknibbs wrote:Have they managed to get text-to-speech engines that actually produce human-sounding speech yet, though? Even satnavs, which you'd think would be a shoo-in for the technology, usually use recorded voice samples, and when they don't, the mispronunciations can be a bit silly.

AFAIK at least some of those those satnavs use pre-recorded snippets that only consist of a few syllable or single words, spoken by a voice actor. Those are then assembled to sentences on the fly.

One real-world example is the cheap satnav I bought a few years ago. Most of the time, it actually gets the pronounciation right. There are a few exceptions, but I still think it would be good enough for Betty, who is supposed to be a machine after all. And if the technology can run in a cheap satnav, it should be easy enough for a modern PC.

Another example I found is https://www.naturalreaders.com/online/, which is about as good as my satnav. It still suffers from inability to recognize context, such as in “No, I’m sorry to say I haven’t.” In this example, NaturalReader pronounces the "haven't" like "haven" in seafaring. But try that one out for yourself, they have a web site you can feed with a few sentences.

Edit:
I should really read the entire thread, someone has mentioned NaturalReader before

.

Alkaiser · Post by **Alkaiser** » Fri, 15. Sep 17, 21:08

For an example of what TTS can do for Betty:
https://forum.egosoft.com/viewtopic.php?t=392407

I made this after I got sick of the inconsistency and cringe-worthiness of some of the Betty audio.

A modern TTS engine is more than sufficient for something like Betty.

If it isn't cost prohibitive, I would be very happy to see Egosoft license a good TTS engine for X4.

RAVEN.myst · Post by **RAVEN.myst** » Sat, 16. Sep 17, 00:43

Alkaiser wrote:If it isn't cost prohibitive, I would be very happy to see Egosoft license a good TTS engine for X4.

This is where I see reciprocal marketing opportunity: a smart TTS provider might offer the license at a discount rate in exchange for brand exposure - now, I don't mean in-game advertising per se, but something subtle such as a small logo or brand name on the corner of the nav console or such - where one might expect to find a commercial brand logo in any case, thus "enhancing immersion via realism/plausibility" (take a look at your monitor, for example - there's a fair chance that there's a manufacturer's logo on a corner or such, though ironically my current main screen doesn't; my older secondary screen DOES bear a Samsung logo, though, and every previous monitor I've owned has shown its bran in a corner of its frame); alternatively, X4's splash screen could bear an acknowledgement (Betty brought to you by such-and-such, or the like), and/or the TTS provider could be given the right to add X4 to its brag-list.

adeine · Post by **adeine** » Sat, 16. Sep 17, 12:00

If this were to happen, the most important thing would be to set up a pretty thorough dictionary for all the X-specific words, and ideally give access to it for modders as well.

A dictionary basically tells the TTS engine "when you see sequence A, pronounce sequence B", so as a hypothetical example if there was a problem with the pronunciation of "xenon" you could do "xenon" -> "zeenon" (or whatever produces the correct phonetic output).

caleb · Post by **caleb** » Sat, 16. Sep 17, 16:28

adeine wrote:If this were to happen, the most important thing would be to set up a pretty thorough dictionary for all the X-specific words, and ideally give access to it for modders as well.

A dictionary basically tells the TTS engine "when you see sequence A, pronounce sequence B", so as a hypothetical example if there was a problem with the pronunciation of "xenon" you could do "xenon" -> "zeenon" (or whatever produces the correct phonetic output).

Oh, good point had not even thought of that... And yeah, I tried that website again:

https://www.naturalreaders.com/online/

With these:

Xenon

Khaak

PAC

HEPT

IRE

And white the Xenon and Khaak sound ok to me, the weapon names are said as a single word. Need to type it as:

P.A.C.

H.E.P.T.

I.R.E.

If I want then spelled out. Good point there. But yeah a dictionary could work that out.