Project Echo - Voice Packs - Deep Learning Voice Synthesis

[Herowynne]

On 7/11/2021 at 10:41 PM, ruddycray said:

I thought I had set it on most of them, but I do see a fair few missing voice types. I'll add it to the list.

 

@ruddycray released version 0.7 of The Ancient Profession, and specified voice types for NPCs who were missing them.

 

Perhaps @Executaball or @judge007 might consider generating a voice pack now? Thanks so much! ❤️

[judge007]

5 hours ago, Nyx Morrigan said:

Could I make a request for Babodialogue and possibly Thief?

BaboDialogue is already done, check @Executaball or @DoubleCheeseburger

 

As for Thief, it is written in German I believe, with various partial translation's none of which were complete.

Last I looked, it had about 5 versions available, duplicated between LE and SE.

 

Along with a multitude of patches, for various issues, and a handful of addons - all of which were at different levels of translation.

 

Complicate that with the fact that Thief provides it's own "Sex engine" and doesn't require SexLab, ZaZ or DD,

and one would have to actually install it, COMPLETELY, in order to get an "accurate" output of voice dialogue for the synthesizer.

 

One would have to do so for each plugin and addon, individually.

 

That wouldn't be the hard part. The hard part would be fielding all the support questions about why 

you included this patch but not that one.

this language and not that one,

this version and not that one

why you haven't finished the translation

 

In short, it's too complicated, to do a job anyone will be happy with in the end.

[judge007]

3 hours ago, Herowynne said:

 

@ruddycray released version 0.7 of The Ancient Profession, and specified voice types for NPCs who were missing them.

 

Perhaps @Executaball or @judge007 might consider generating a voice pack now? Thanks so much! ❤️

 

I've had quite a busy time of it lately.

I have a couple of mods I wish to do.

But I need to take a breather and analyze the results I've had.

 

I made a major mistake with SexLab Dirty Talk, by not checking it's internals.

It was never written with voice in mind, doesn't assign voice models to anything or anyone.

It assigns it's sub-title text to anyone in the cell, and never relinquishes it.

While the synthesizer output, it gets distributed randomly.

 

I don't wish a repeat of that.

 

When I get my Load order repaired, I may actually play a bit as well.

 

Not saying no, just not in the near future.

[judge007]

On 7/18/2021 at 9:14 AM, judge007 said:

Working with the LoversLab file size limitations, I'm running into issues with multipart archives appearing to work,

but having no sound after recombing and installing.

 

Once that issue is fixed, I have several I'll post.

I know the esl count is dependant on the number of FormIDs.

With the dummy esl's we're putting out for the voice packs, there are no FormIDs, so theoretically aren't an issue.

However trying to post SexLab Dirty Talk in Multipart archive (3 bsa's, 3 esl's) for some reason get no sound.

Single BSA/ESL here functions, multipart not.

Looking at my install I have many BSA's,

But like you can find no reference as to limitations.

Could also be some form of compression issue

 

The issue was SexLab Dirty Talk itself.

The mod contains just enough data to have the Synth script operate,

but was never written with voice in mind.

Original author was working towards a "lite' version of Apropos 2, without the complex database of text.

Mod author put in enough Dialog data to get text to Sub-titles and stopped there.

No voice model was specified, or active author, so results with speech are random.

 

If anyone wants to pickup that project, feel free, as well as for-warned.

 

[Executaball]

6 hours ago, Nyx Morrigan said:

Could I make a request for Babodialogue and possibly Thief?

 

Thief should be okay, I'll take a look.

 

@DoubleCheeseburger maintains a voicepack for Babodialogue v3.901 here:

 

Though the current public release is v4.02, I don't think much dialogue has changed since then, mostly bug fixes.

[Executaball]

5 hours ago, Herowynne said:

 

@ruddycray released version 0.7 of The Ancient Profession, and specified voice types for NPCs who were missing them.

 

Perhaps @Executaball or @judge007 might consider generating a voice pack now? Thanks so much! ❤️

 

Great. I think my only problem with it was the voice types, so it should be okay now. I'll add it to the list~

[Executaball]

WIP on SynthUtil to process csv's for word replacements and format fixes for use with XVASynth. 

 

Next after doing some more voice packs I will work on phoneme sequencing and interfacing directly with FaceFXWrapper to generate .lip files without Creation Kit. At this point, so far, it seems possible. But we'll see. 

Edited July 23, 2021 by Executaball

[Yuni]

Ooooh! Public Whore sexlab sound samples! Can we get a full version of that? I love that mod, hearing people say that making me naked was the best decision the Jarl ever made, would be great! Babo's Dialogue (aka BakaFactory's dialogue) would be great too.

Thanks for all the hard work deary!

[judge007]

11 hours ago, Yuni said:

Ooooh! Public Whore sexlab sound samples! Can we get a full version of that? I love that mod, hearing people say that making me naked was the best decision the Jarl ever made, would be great! Babo's Dialogue (aka BakaFactory's dialogue) would be great too.

Thanks for all the hard work deary!

Go to @Executaball's first page, it's there

 

EDIT:

correcting user name, is there some form of auto-correct running here?

 

Edited July 24, 2021 by judge007

[Executaball]

40 minutes ago, Yuni said:

Ooooh! Public Whore sexlab sound samples! Can we get a full version of that? I love that mod, hearing people say that making me naked was the best decision the Jarl ever made, would be great! Babo's Dialogue (aka BakaFactory's dialogue) would be great too.

Thanks for all the hard work deary!

 

The actual sound pack for Public Whore is out and on the page, just scroll down.

[spoonsinger]

Skoma Whore probably needs the treatment. (Unless my googlefu is rubbish, or my searches are being censored - but seriously that would quite nice for a future LL based load out).

[Executaball]

On 7/22/2021 at 7:13 AM, Herowynne said:

 

@ruddycray released version 0.7 of The Ancient Profession, and specified voice types for NPCs who were missing them.

 

Perhaps @Executaball or @judge007 might consider generating a voice pack now? Thanks so much! ❤️

 

Ancient Profession v0.7 completed and uploaded~

 

I don't play the mod that much myself, so let me know how it goes ^^ 

[Herowynne]

57 minutes ago, Executaball said:

Ancient Profession v0.7 completed and uploaded~

 

I don't play the mod that much myself, so let me know how it goes

 

Thank you so much! ? Yes, I will report back how my testing goes.

[AxeMEMES]

IDK if anyone has requested you do one for the mod "Sexlab Eager NPCS" (SLEN) or "Arousal Based Breast Adjuster" (ABBA) 
those two mods would do nicely with this system 

Edited July 24, 2021 by AxeMEMES

[judge007]

13 minutes ago, AxeMEMES said:

IDK if anyone has requested you do one for the mod "Sexlab Eager NPCS" (SLEN) or "Arousal Based Breast Adjuster" (ABBA) 
those two mods would do nicely with this system 

SLEN is on @DoubleCheeseburger's thread https://www.loverslab.com/topic/168547-doublecheeseburgers-xvasynth-based-mod-voicification-thread

 

[Executaball]

18 minutes ago, AxeMEMES said:

 "Arousal Based Breast Adjuster" (ABBA) 

 

How much dialogue does that mod have? o.o

[judge007]

@Executaball & @DoubleCheeseburger

 

I'm currently about a third of the way through Sexist Guards.

Just checking to see if either of you have already done it or were planning to.

 

No need for duplicated efforts

[Herowynne]

8 hours ago, Herowynne said:

Yes, I will report back how my testing goes.

 

@Executaball So far, so good! I tested all the dialogue options with the Whiterun carriage driver, Bjorlam.

 

Overall, the quality of the voiced dialogue was excellent! The only klunky parts were where Bjorlam tries to pronounce the acronym "S.L.U.T.S." ?

Edited July 24, 2021 by Herowynne

[judge007]

3 minutes ago, Herowynne said:

 

@Executaball So far, so good! I tested all the dialogue options with the Whiterun carriage driver, Bjorlam.

 

Overall, the quality of the voiced dialogue was excellent! The only klunky parts were where Bjorlam tries to pronounce the acronym "S.L.U.T.S." ?

 

Sounds like more my voice pack S.L.U.T.S. than executaball's, but good to get feed-back regardless. Thanks.

[Herowynne]

1 hour ago, judge007 said:

Sounds like more my voice pack S.L.U.T.S. than executaball's, but good to get feed-back regardless. Thanks.

 

Yes, I feel really silly. I got my feedback mixed up for the different voice packs I have recently installed.

 

You are correct, @judge007 That feedback was meant for your S.L.U.T.S voice pack, which is great!

 

@Executaball I will also post my feedback for The Ancient Profession voice pack. Thanks again!

[Herowynne]

1 hour ago, Herowynne said:

I will also post my feedback for The Ancient Profession voice pack.

 

@Executaball Started a new game with The Ancient Profession 0.7 and your voice pack. Tested all the dialogue options with the initial conversation with the madam of the Solitude brothel - it all sounded really good.

 

The Ancient Profession has a lot of dialogue, and I feel that it really benefits from being voiced. Much more immersive!

 

I will play through the quests again and post more feedback here. Many thanks!

[Executaball]

Initial indexing for Thief shows at least 370,000 unique voice lines. This is not including player voices which is handled by the Dogma Engine. Including player voices (which I have assigned a single voice type, FemaleEvenToned) this comes out to be 381,708 lines. This will be the largest mod I have ever attempted.

Edited July 25, 2021 by Executaball

[judge007]

26 minutes ago, Executaball said:

Initial indexing for Thief shows at least 370,000 unique voice lines. This is not including player voices which is handled by the Dogma Engine. Including player voices (which I have assigned a single voice type, FemaleEvenToned) this comes out to be 381,708 lines. This will be the largest mod I have ever attempted.

 

How much remains untranslated?

[Executaball]

9 hours ago, judge007 said:

 

How much remains untranslated?

 

All lines translated, using the BME patch for v1.0.0.8.

 

 

[Executaball]

On 7/22/2021 at 5:17 AM, Nyx Morrigan said:

possibly Thief?

 

In Progress~

 

Samples attached  

 

FemaleCommander_bd_MQ__000AE3A9_2.wav FemaleCommander_bd_MQ__000AE3AB_1.wav FemaleCommander_bd_MQ__000AE3A9_1.wav MaleEvenTonedAccented_bd_MQ__000A92B0_1.wav MaleEvenTonedAccented_bd_MQ__000A92B1_1.wav