Additional TTS Voices for Windows

[Exigeous]

I've done a great deal of searching so if I've missed something my apologies and thanks for a link.

I'm looking for some add-on text to speech (TTS) voices for Windows 10.  While I enjoy the main Voice Attack voice I've built many custom actions for my setup and use the built in female Windows voice for feedback.  It works just fine but I don't really love the voice nor does it match the pre-recorded VA voices.

I'm a big fan of more robotic voices such as the voice in No Man's Sky or Crysis 2.  That very digital buzzy type voice.  As the current Windows voices don't sound that human anyway I'd love to just use a robotic voice and embrace the fact that it's a computer speaking back to me.

Anyone know of a solution or way to accomplish this??

-Ross

[Pfeil]

Good text to speech voices are generally expensive, both to make and to buy as an end user, so novelty voices that are still understandable will likely be hard to come by.

For voices in general, the highest quality ones I've found are made by Ivona. They're not cheap, but they're the closest to natural I've encountered.


I've Googled around a bit, but it doesn't look like there's a way to do live(realtime) effects on SAPI voices.

This post on the old VoiceAttack Google Group suggests capturing the audio that TTS produces and applying the effect to the recording is theoretically possible, but that requires implementation of an entirely separate library into VoiceAttack to make it work.

[Exigeous]

Thanks for the great information and links.  Looking at Ivona's site they have SAPI 5 voices for $45 - am I understanding correctly that those voices would be installed and then available in Windows 10's TTS settings?  The Sali voice is pretty great with only minor issues letting you know it's TTS.  While I would love to find something stylize if that's not really possible then I'd consider a better TTS over what comes with Windows 10.

I'm also wondering about your final comment, if these could somehow be routed through a processing filter to add effects.  I use a an audio routing app called VoiceMeter, I'm wondering if that could take the input from TTS and route it to another app to add the effects.  Suppose I'll play when I get home.

[Antaniserse]

Thanks for the great information and links.  Looking at Ivona's site they have SAPI 5 voices for $45 - am I understanding correctly that those voices would be installed and then available in Windows 10's TTS settings?

Yes, they do... they'll show up in the OS Speech control panel, and you can use them with any TTS application; I had occasion to try Ivona Amy with VoiceAttack and it sounds fantastic

Actually, I think all 3 British voices sounds great and natural, don't know if it's a matter of accent or what, but it fits TTS almost perfectly

[Exigeous]

I definitely agree with you on the British voices, they real do sound the best.  It also seems that female voices are far better than male.  I don't think that because I'm a man and would want a woman, I just think they objectively are better.

Thanks for confirming about SAPI 5 voices.  While $45 isn't the bad at the same time I'm not sure it's worth $45 to make the dozen or so custom commands I've created sound a bit better.  I actually speed up the built in Windows female voice (forget the name) by 1, I feel like that slight speed increase goes a long way toward realism.

Not sure if you're familiar with Elite Dangerous and by extension Radio Sidewinder.  It's a streaming station that plays music provided by the community (that fits with Elite, no Country!).  What's great about it is every so often they play a news segment and fake product ads.  They do a really fantastic job with it and, for me, it just adds on to the immersion.  I bring this up as one of their newscaster is "Robot Rachel".  I didn't realize it was a TTS voice until someone pointed it out during a Discord chat.  Now that I know I can tell but it's extremely well done.

Okay, now to think on the SAPI 5 voice.  It's too bad there's not some kind of limited time/function demo, if I could hear it in flight I'd know if it's what I want or not.

[Antaniserse]

Thanks for confirming about SAPI 5 voices.  While $45 isn't the bad at the same time I'm not sure it's worth $45 to make the dozen or so custom commands I've created sound a bit better.

Well, if you just need this limited number of entries, and they are fixed, you can cheat a bit:

if you run a software like Audacity and such, you can grab your sound card output directly to a WAV file, so you may go to the Ivona main page, select your desired voice, input your phrase in the demo text below and press Play --> voilà, you have a sound sample to use in VA.
Rinse and repeat

Okay, now to think on the SAPI 5 voice.  It's too bad there's not some kind of limited time/function demo, if I could hear it in flight I'd know if it's what I want or not.

Actually, there used to be one, 30 days limited... I'm 100% sure I used that in the past, but I can't seem to find any download anymore on the main site
However, one of the official distributors still seems to have it, if you enter in the details page of any of those, there is a download available

[TheThingIs]

I've Googled around a bit, but it doesn't look like there's a way to do live(realtime) effects on SAPI voices.

Pfeil, effects can be done live. Check out EDDI for Elite as they manage to do it, I have no idea how it's done though

[Pfeil]

Pfeil, effects can be done live. Check out EDDI for Elite as they manage to do it, I have no idea how it's done though

From what I can see in the source code, it's done as previously stated:

capturing the audio that TTS produces and applying the effect to the recording

The output of the TTS synthesizer is stored in memory, effects are applied, and the data is then played back as sound.

I would argue a filter, on the other hand, applies the effect as the sound passes through, directly to each individual part of it, rather than as a whole.

From a users' perspective, the semantics don't matter much, as synthesized speech is produced nearly instantly(by human perception at least) rather than having to wait for text to be read out, so the process of applying the effect and playing the result shouldn't add a noticeable delay, making it appear realtime.

[TheThingIs]

ahhhh, so that's how it's being done!

Maybe Gary could do something similar in va and give us some options to add different kinds of effects to any sound file, which would also include a tts as it's converted to a sound file before adding an effect.

[Solracm]

I have a fully legit copy of Ivona Amy I bought as unwanted gift from Deskshare.com. I have the exe and email with unused licence key.

It cost around £25. I will accept £15 ono. PM me for more details.

Carlos

[ralf44]

Going back to the OP's objective,
"I'm a big fan of more robotic voices such as the voice in No Man's Sky or Crysis 2. "

I am doing exactly that in realtime by funneling sound output from VA's TTS thru Facerig, which provides lipsynched avatar animation but also live voice effects, including a robotic one. This is cheap but not free software, best bought on Steam.

However if you are lucky enough to have Intel's new Realsense technology, either built-in to your computer or as a standalone sensor, there is a free version of Facerig built for that which does live face and hand puppetry on top of the lipsynch:
https://facerig.com/irfacerig/

(Intel Realsense is basically Xbox Kinect with higher quality, for Windows, and is being resold by Razer for gaming and by Microsoft to push Windows Hello.)

[mikelimtw]

I've done a great deal of searching so if I've missed something my apologies and thanks for a link.

I'm looking for some add-on text to speech (TTS) voices for Windows 10.  While I enjoy the main Voice Attack voice I've built many custom actions for my setup and use the built in female Windows voice for feedback.  It works just fine but I don't really love the voice nor does it match the pre-recorded VA voices.

I'm a big fan of more robotic voices such as the voice in No Man's Sky or Crysis 2.  That very digital buzzy type voice.  As the current Windows voices don't sound that human anyway I'd love to just use a robotic voice and embrace the fact that it's a computer speaking back to me.

Anyone know of a solution or way to accomplish this??

-Ross

Is this for general Windows use or for Elite Dangerous?

[ralf44]

"Is this for general Windows use or for Elite Dangerous?"

Shouldn't make any difference...

[Exigeous]

First I did actually buy the Emma voice from Ivona.  I have to say it's rather amazing, most of the time I can't tell it's not human and I'm the one that installed it!  When streaming on Twitch I routinely get questions about it, "what's that speaking?" as they know VA/EDDI but don't understand how I have a human voice reading it.  The wow/immersion factor is rather high.

I am doing exactly that in realtime by funneling sound output from VA's TTS thru Facerig

Thanks for sharing that as it does look rather cool.  I watched the trailers/demos on the store page but didn't see that it did audio effects, just the face mapping.  Did it just not share that?  Do you know if you can do the audio effect without the video, meaning I could use my Rift microphone (which is rather great) and take it's output and pipe it into Facerig?  Also what about piping a single application through it, for example both Elite and EDDI/VoiceAttack use the standard 5.1 card on the motherboard.  I'd want EDDI/VA to get the processing but not Elite.  Would I need to do some trickery with something like VoiceMeeter with virtual in/out?

"Is this for general Windows use or for Elite Dangerous?"..........Shouldn't make any difference...

For this use case Elite - but really for VoiceAttack/EDDI - really anything spoken with TTS.  I want Elite's audio to stay normal, I'd turn off all the voices there and then have VA/EDDI "say" everything via TTS which would pipe into this to add the "robot" effect I'm looking for.

[ralf44]

Necro-posting because this has a core idea of wider interest. We were discussing adding a realtime postprocessing effect to VA's text-to-speech output, for theatrical effects, such as octave shift, robot voice, and more advanced effects like a Granulator are feasible using freeware.

Facerig is not actually the best solution, I just mentioned that it can do ie robot vocoder effect in realtime - as can some apps aimed at videochat or livestreaming, but if you're not using the furry 3d animated avatar then you're taxing your graphics card for no reason when it's already busy - and they have some commercial licensing restrictions that could possibly affect you.

We actually want a lightweight app that just takes a selected Windows Audio Device as input and applies one or more VSTs of your choice to it and pipes it to a selected Windows Audio Out device, real or actual, with the shortest delay possible and least overheads.

I think "VST" is the magic word to research on, it's an old plugin standard for live audio processing and there are a lot of options and a million free filters to choose from.

Maybe an engineer or programmer can suggest something barebones that would suit streaming, I do not have a go-to version of this on Windows. But Facerig can absolutely do it, as one option that I've tested.

It would also be possible for Gary to go down to his shed and bang VST support into VA, in theory :p

[Gangrel]

Voicemeter Banana can accept VST plugins.... go from there and the routing of audio output paths and the like

[ralf44]

https://www.vb-audio.com/Voicemeeter/banana.htm (Donationware)

ty Gangrel

Same dev makes Virtual Audio Cable which I've used for streaming in the past:

https://www.vb-audio.com/Cable/index.htm