Originally posted by Lunrei:

Have you tip or clue to find an other voice?
I'm sick of Hortense. Even if Julie is install and selected it don"t appear in MTW.

Do you know where i can find a voice (i don't care if is free or not), i finded several website, but they sell their voice with a reader. (not even compatible with win11)
How can i use a file if i get one in MTW.

Hi, Julie does not appear because it is a 32-bit voice, MTW only recognizes 64-bit voices.

https://www.cereproc.com/en/storesapi

This page looks good, but I haven't tried it. It says that they are compatible with Microsoft SAPI 5 voices, the API implemented in MTW. If the voice is 64-bit and is installed in the default Windows folder (which would be logical, otherwise you would have to copy it there manually. ), MTW should recognize it in the settings menu.

Thanks a lot for your advice.
I took one their voice (Suzanne) and installed it and it work with their reader. So , we have some problems.

If i select MTW to be in english, only Zira appear in the setting list and then in game it is again the same and only voice who is use.

If i select german. Only Zira appear in the setting list., and in game is the same. (i have no german voice install in my windows.) Zira is again the only voice used and available.

Now if i select French, i have Hortense, Zira, Suzanne, Haruka, Haemi, Huihui, Hanhan.
but In my windows11, i have much more:
for french : Hortense, Caroline, Julie, Paul, Claude, Guillaume, Suzanne.
I have also Tamul, chinese simplified (Huihui) and Taiwan (Hanhan), japannese (Haruka), Korean (Heami)

So if MTW is in french and i selected Haruka. then in the game until Marian speak it is Hortense who is selected by défault, but i can choose at this moment between Hortense, Suzanne and Haruka.
if i select again Haruka, it work, it spoke to me in japanese accent. but if i select Suzanne, it is the voice of Hortense who speak instead.
Suzanne and Hortense always appear in the list when i choosen french, And it is the same if i selected Suzanne at first, of course. But It don"t work as the same.

I hope it is clear, not easy to explain without screenshoot i wished to show you.

Why only some voice are present in MTW? of course i don't use all of them, it was for a test.
For french voice, i have only Hortense who is used and available in MTW, whatever voice i have select for windows11.
My concern is for Suzanne of course.
note: Suzanne don't appear in the list of windows voices.

Ok, please open windows registry editor.

https://www.youtube.com/watch?v=ZQTwM9X-ckU&ab_channel=TechProAdvice

go to this folder:

Computer\HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech\Voices\Tokens

Is Suzanne there?

If she is there, could you please tell me what is: Gender, Labguage, Name and Vendor.

Gender = Female
Langage = 40C
Langage_ID = fr
Name = CereVoice Suzanne - French (France)
Vendor = CereProc Ltd
VendorPreferred = **it is empty**

Aha, it has the same code (language), as I thought, Hortense is also 40C. Ok, excellent, then it is there.

Within MTW the voice execution is done with the language and gender attributes only, that is why it selects Hortense, Windows gives it preference. That means that for the next update I must also add the name as an attribute.

For now, what you can do is uninstall Hortense (windows settings) so that Suzanne takes its place, after the next update you will be able to reinstall Hortense and have both available.

As i said , i'm not thrill to have Hortense be back... lol

Not sure i can uninstall her, She was with Win11 installation by default. On the front, win11 don't allow me to delete it.
I could remove her manualy i beleive, but i don't want screw up my key register if i do wrong.
I don't know enought to manipulate such thing.... I will wait. :)

Ok, good, Thanks for the feedback to make MTW better!

Also thanks to you for your help, i will report on your post what i think of Cereproc's voice, once i can use it in MTW. ;)
I have some concern... but i want to see. To hear i should say.

Originally posted by Lunrei:

Also thanks to you for your help, i will report on your post what i think of Cereproc's voice, once i can use it in MTW. ;)
I have some concern... but i want to see. To hear i should say.

Hi Lunrei, with the new update, now the external voice should works, let me know, please!

So yes, it work! Hortense can steel be selected once in game (in front of Marian) But Cerevoice Suzanne can also be use now.

My concern has unfortunatly hit something.
Cerevoice have indeed good natural voice, but don't fall with their trick.

If you try on their website a text (they don't hide it, they tell you an IA analyse your text to add variable to be use by a voice of your choice) like this:
La maison de Lunjei<break type="2" time="0.1" /> a de grande pièces pour vivre,<break type="4" time="0.6" /> un bureau, un salon, une salle de bain, une chambre à<break type="2" time="0.1" /> coucher, une cuisine<break type="2" time="0.1" /> et des couloirs,<break type="4" time="0.6" /> le style est moderne en verre<break type="2" time="0.1" /> ou métal<break type="2" time="0.1" /> relever par une décoration sobre. Les pièces sont clair et possèdent de grande fenêtre.

It add pauses and fluctuations, lower or higher pitch.

The voice with their reader can have a very realistic voice. At least as close as you can dream of.
However without these variables, you are more close to Hortense or Zira. ^^
Roughtly the voice is slightly better than a basic one, but not as far as you can hope.

Ah, so that's the trick. Those are Microsoft SAPI commands.

I think that while applying an AI technique would be too much, we could apply some simple rules to make Marian's way of speaking a little more real.

The information I do have is regarding when she asks (the sentence ends in ?), exclaims (ends in !) or, let's say, is normal (ends in . ). For the first two, we would have a average configuration for the question tone and the exclamation tone. For normal sentences, I can think of at least putting in a small random fluctuation to break the monotony but without being too much so as to fall into chaos.

Please, when you can, could you put something like this (in French) to know what values ​​the AI ​​puts:

What is your name?
Do you like being with me?
How are you?
Do you trust me?
You're amazing!
Get naked!
Let's play!
Congratulations!
I feel sad.
I feel very happy.
It's a very boring day.

Originally posted by JORGE:

... Please, when you can, could you put something like this (in French) to know what values ​​the AI ​​puts:
What is your name? ...

I'm not sure to understand what you mean. And why in French? (each spelling have is own way to "breathe") What is your name and Quel est ton nom? Don't expel the same way your air from your lungs if you see what i mean.)

My little text was an exemple i maded my self quickly. When you buy a voice on Cerevoice website, you have also a softwear/reader to test manualy these commands. I don't have enought inside's credit to use again their website for now. i must wait 10 days yet to "try" an other text.
But in my reader i have these commands i can add to my text manualy. if it is what you want?

Variant tag <usel variant="1"></usel>
Phones tag <phoneme alphabet="ipa" ph=""></phoneme>
Pitch tag <prosody pitch="+0Hz"></prosody>
Contour tag <prosody contour="(0%,+0Hz) (50%,+10%) (75%,+20Hz)"></prosody>
Rate tag <prosody rate="100%"></prosody>
Volume tag <prosody volume="+0.0dB"></prosody>
Emphasis tag <emphasis level="moderate"></emphasis>
Sentence break tag <break type="4" time="0.6" />
Phrase break tag <break type="3" time="0.3" />
Short break tag <break type="2" time="0.1" />
Happy emotion tag <voice emotion="happy"></voice>
Sad emotion tag <voice emotion="sad"></voice>
Calm emottion tag <voice emotion="calm"></voice>
Cross emotion tag <voice emotion="cross"></voice>

Related to Suzanne only, i have also "vocal gesture". I beleive each voice must have their own list.
<spurt audio='g0001_001'>tut</spurt>
<spurt audio='g0001_002'>tut tut</spurt>
<spurt audio='g0001_003'>cough</spurt>
<spurt audio='g0001_004'>cough</spurt>
<spurt audio='g0001_005'>cough</spurt>
<spurt audio='g0001_006'>clear throat</spurt>
<spurt audio='g0001_007'>breath in</spurt>
<spurt audio='g0001_008'>sharp intake of breath</spurt>
<spurt audio='g0001_009'>breath in through teeth</spurt>
<spurt audio='g0001_010'>sigh happy</spurt>
<spurt audio='g0001_011'>sigh sad</spurt>
<spurt audio='g0001_012'>hmm question</spurt>
<spurt audio='g0001_013'>hmm yes</spurt>
<spurt audio='g0001_014'>hmm thinking</spurt>
<spurt audio='g0001_015'>umm</spurt>
<spurt audio='g0001_016'>umm</spurt>
<spurt audio='g0001_017'>erm</spurt>
<spurt audio='g0001_018'>err</spurt>
<spurt audio='g0001_019'>giggle</spurt>
<spurt audio='g0001_020'>giggle</spurt>
<spurt audio='g0001_021'>laugh</spurt>
<spurt audio='g0001_022'>laugh</spurt>
<spurt audio='g0001_023'>laugh</spurt>
<spurt audio='g0001_024'>laugh</spurt>
<spurt audio='g0001_025'>ah positive</spurt>
<spurt audio='g0001_026'>ah negative</spurt>
<spurt audio='g0001_027'>oui question</spurt>
<spurt audio='g0001_028'>oui positive</spurt>
<spurt audio='g0001_029'>oui resigned</spurt>
<spurt audio='g0001_030'>sniff</spurt>
<spurt audio='g0001_031'>sniff</spurt>
<spurt audio='g0001_032'>argh</spurt>
<spurt audio='g0001_033'>argh</spurt>
<spurt audio='g0001_034'>ugh</spurt>
<spurt audio='g0001_035'>ocht</spurt>
<spurt audio='g0001_036'>ouais !</spurt>
<spurt audio='g0001_037'>oh positive</spurt>
<spurt audio='g0001_038'>oh negative</spurt>
<spurt audio='g0001_039'>sarcastic noise</spurt>
<spurt audio='g0001_040'>yawn</spurt>
<spurt audio='g0001_041'>yawn</spurt>
<spurt audio='g0001_042'>snore</spurt>
<spurt audio='g0001_043'>snore phew</spurt>
<spurt audio='g0001_044'>zzz</spurt>
<spurt audio='g0001_047'>brrr (froid)</spurt>
<spurt audio='g0001_048'>snort</spurt>
<spurt audio='g0001_050'>ha ha (sarcastic)</spurt>
<spurt audio='g0001_051'>doh</spurt>
<spurt audio='g0001_052'>gasp</spurt>

Enought funny some of them. ^^

What you sent me is very useful. Basically my idea is to have three cases: configuration in question, in exclamation and normal. I need to test if my idea makes sense or if the execution is garbage, we'll see! Thank you very much!

i have notice Cerevoice Suzanne react to "!" and "?" I hear different pitch with the last word.
But not every times, or i don't pay attention enought, i don't understand yet.
Some basic emotion have to be in her voice "core" if i can speak like that.
As i said, the basic cerevoice perform above than Zira or Hortense, but not like on their website.

I have run the tests, but I didn't get a good result :(, with the settings I tried it sounds worse than the default. No way around it, at least for now the commercial voices can be performed, thanks a lot for the help!