Modify TTS Voice to be more "sexy"

[Ectropian]

I wasn't sure exactly what sub to put this in, so I thought this might be good.

So I watch some Twitch, and if you've ever seen Andy with GamingForFun I'm sure you know his "AI" voice is probably the sexiest ever.. I guess you could say it's more "natural" sounding.  So I'm trying to learn a bit more about modifying the voice inputs to VA.  I'm currently using Ivona's Salli (https://harposoftware.com/en/english-usa/129-salli-american-english-voice.html) which is awesome!!  But in VA she's very choppy and abrupt.

I'm also using EDDI, which allows some processing levels to be modified as well as speed of speech.  Does anyone know of any other software that will let you adjust the voice permenantly, or even a legend of speech modification "tips and tricks" to make the voice more natural?

[Exergist]

I'm guessing you're aware of the existing Volume and Rate controls at the bottom of the 'Say Something with Text-To-Speech' panel.

Adding commas, colons, hyphens, and periods introduces pauses in a spoken text phrase. So maybe strategic placement of those types of punctuation may help a bit? By my uncalibrated ears it seems like there is a bit more pause with periods than the other types, but maybe it's my imagination

[Gtrshop]

I've listened to several of the voice packages, and one of the better ones is MS Hazel, noticeable UK accent, sounds much better than most, IMO.

[llama]

I've installed various TTS voices from Microsoft, and I can select and use them from the Windows TTS control panel, and I've set Hazel to be the system default voice.

VoiceAttack displays these voices in the Voice pull-down menu. Good! Unfortunately, no matter which entry in the menu I choose, including selecting Hazel directly, actually changes the voice VA uses for TTS.

I'm in Windows 7.

Any thoughts?

[Exergist]

So you're saying that the voice options make no difference?

As an example, here's what I have in stock W10:


What is the voice that gets used?

[Mike308]

The other path is through something called SSML, Speech Synthesis Markup Language. In particular, it has a variable called "prosody" that places, hm, a bit more emphasis on words until you end prosody. It is not all that difficult but it can be a bit picky in setup, notably if you are using say a UK-Englich voice versus an American English voice. But if you really want to take the time to tweak your TTS strings, you can do a lot to improve the natural sound.

I think that the slight mechanical reverb effect you can generate with the EDDI plug-in also smooths over some of the tiny 'nicks' in Windows TTS speech, even if you are using a quality voice like an Ivona. But speech processing is apparently a bit complex to program and at present I don't know of any plans to bring that to VA outside of EDDI.

I made a post on SSML a while back if it is any value to you:  https://forum.voiceattack.com/smf/index.php?topic=1659.0