Text to Speech voices - the realistic ones

Archives Forums/General Discussion/Text to Speech voices - the realistic ones

puki(Posted 2007) [#1]
Everybody knows about Microsoft "Sam", "Mike" and "Mary", etc. Yet there are some very realistic Text To Speech engines out there.

I just installed the Microsoft Text-to-Speech Engines for SAPI5:

This comes with a sample TTS voice - however, there is no info about it - it is chuffingly realistic.

I note from a quick Google about that you can download these voices for around $30 USD / 15.00 UK:

Some of the voices you can listen to on that page are amazing.

Anyone know much about TTS voices - any really great ones to go for (each for circa the price of a computer game)?

big10p(Posted 2007) [#2]
Are you still trying to build a virtual friend?

puki(Posted 2007) [#3]

Now I can have loads of them.

Azathoth(Posted 2007) [#4]
Will they be even more awesome than you?

puki(Posted 2007) [#5]
The thing is, I need them to work with Microsoft Speech - then I can use them in Blitz3D.

I got the sample, high quality, one to work.

I don't know enough about TTS. I am hoping that as long as they are SAPI5 compliant, they will work.

jfk EO-11110(Posted 2007) [#6]
Thought sapi 5 i way outdated, no?

When I used it last time, it was pretty crappy IMHO, compared to state of the art speech.

There are two modes in sapi 5, one is based on siblings and sounds relative synthetic. The other one is based on a lexical words list of samples. They got one example included that sounds pretty good, no surprise if you simply record the phrase word by word. I then tried to create new voices based on this more simple way of adding voices to the speech api, but it didn't work for some reason.

I think XP has something newer than SAPI5, no?

puki(Posted 2007) [#7]
The SAPI5 are certainly not as good as some of the other ones that are 'TextAloud' compatible. However, I think that those ones will not work in Blitz3D.

I just installed the Cepstral Character Whispery voice.

Works fine in Blitz3D - It was $6.99 USD - so it cost me about 3.50.

You can download and install demo versions: The trial period is not time-limited; It never expires. However, a voice will interject reminder messages into the audio output until you purchase a valid license key for it.

Will probably require a reboot to work with Blitz3D, but should work in Windows Speech straight away.

You can listen to the voices here:

xMicky(Posted 2007) [#8]
(each for circa the price of a computer game)

I don't know whether you want to make this virtual friend-project only for yourselve or to share it whith others. Only if last alternative is true -
are you sure you are allowed for that price to distribute the generated sound files as a part of a commercial or even a non-commercial product ?

When I click on "Site licences" on's site, I find that they speak only of a use related to a special computer you bought the software for. From that I would guess they don't allow distributing .wav-files made with their software. May be the complete license text shipped with the product tells something different...

puki(Posted 2007) [#9]
I won't be distributing.

However, The TTS sample voice that comes with the MS SAPI5 installation is high quality - Unfortunately, MS failed to mention what it is and where you can buy the full version. However, it does work in Blitz3D.

John Blackledge(Posted 2007) [#10]
Puki, since I've totally failed to get text to speech working ever (XP tells me it's enabled but I hear nothing) is there any chance you could knock together a quick and dirty demo of what you're doing?

puki(Posted 2007) [#11]
Quite often only 1 voice is installed as standard.

You need to have the 'Speech' icon in your control panel - it isn't always there.

I'd recommend downloading and installing 'Microsoft Text-to-Speech Engines for SAPI5' - as in my first post.

This is a big-boy installation and is a complete installation of SAPI5.

The Blitz3D part of this relies on you having speaker.dll and speaker.decs

Grey Alien(Posted 2007) [#12]
Are you going to combine a tiny CPU and sound system with this software in one of your real dolls, and then program in some Puki-loving AI?

puki(Posted 2007) [#13]
I've uploaded 'So-to-Speak' for Blitz3D:

Once downloaded - change the file extension from .wmv to .zip.

I had to alter the extension to get the upload to work - it's a normal archive.

puki(Posted 2007) [#14]
Mmm, that proved not a good idea.

I've battered my way into their servers and decoded the link:

Not sure how long it will be before they spot this and kill me.

puki(Posted 2007) [#15]
Note: clicking that link may not work - but you can paste it into your browser address bar and it will kick into their server.

puki(Posted 2007) [#16]
Mmm, try this one:

Actually, the thing is moving - they obviously don't have static links to files - it seems to move each time I try to access it - or, it is redirecting to different file servers.


The only way you can probably grab it is to click:

Then right-click the main view window and select properties.

Copy and paste the Location into your browser.

You may only be able to get the Location via Firefox.

Having said that, if you only have IE and you cannot right-click to get the properties, then right-click away from the view window and select 'view source' - then CTRL Find 'wmv' - the link is there.

It would have been easier if I just uploaded it to a file server - but I like Putfile.

It is easier to do with FireFox.

jfk EO-11110(Posted 2007) [#17]
I guess video space is not the only space you need these days.

WendellM(Posted 2007) [#18]
I always enjoy the unpredictable new directions you get interested in, man, and your unique way of exploring them.

skidracer(Posted 2007) [#19]
AT&T have the tech, I'm not sure this demo page does them justice:

AdrianT(Posted 2007) [#20]
My GPS uses TTS voices, its quite funny. I have a male and female british english speakers, and american accented ones too.

The guy sounds a bit like a cross between prince charles and a radio presenter lol. The woman sounds like someone from BBC news. The american woman sounds like a waitress from somewhere midwest. Can't remember what the american guy sounds like.

Wonder if I can coppy the TTS files to my PC and use them in windows.

John Blackledge(Posted 2007) [#21]
Ah right.
(so apart from the fact that I can't access your file) this needs the user to have already installed MS Text to Speech.
Nah, most of them can just about double-click a setup.exe.

Hotcakes(Posted 2007) [#22]
I think running that thing on Vista replaces the default TTS engine of 8.0 with 5.1.

The default MS voices that come with Vista no longer work, but Anne is far superior anyway. Also that sample one is just a bunch of samples. If you get it to say anything other than the default test string, it will replace every unrecognised word with 'blah'.

_33(Posted 2007) [#23]
Are there ANY examples in blitz on using a text to speech?

puki(Posted 2007) [#24]
'So to Speak' does it:

That is what I am using.

semar(Posted 2007) [#25]
Are there ANY examples in blitz on using a text to speech?

SoToSpeak is written in BlitzPlus and uses a DLL made from Metalman. You can download SoToSpeak exe and its complete source code + DLL at my website:
under the section: Projects.


puki(Posted 2007) [#26]
That's the sausage. That's the thing I tried to upload.

It works fine for me.

jfk EO-11110(Posted 2007) [#27]
I think there are several speech userlib examples in the code archives.

John Blackledge(Posted 2007) [#28]
Well, I'm delighted. Thanks Puki.

My PC has a fairly new, clean reinstall of XP (+SP2).
Even so the download from sergio worked immediately without having to install the Microsoft Text-to-Speech Engines for SAPI5. Which is what I think you were getting at.
(Yes there is a Blitz-Plus exe, but also a BlitzBasic .bb example.)

Much fun to be had with this.
Keep us posted on what you are doing with this.

John Blackledge(Posted 2007) [#29]
Speaking the ReadMe.txt file:-
(Copy the decls to 'userlibs', leave the dll where you downloaded it. Save this file into the download folder, then run it.)

Hee-hee. I'm just a big kid and will never criticise Puki again.

puki(Posted 2007) [#30]
I am the greatest.

John Blackledge(Posted 2007) [#31]
Damn. What did I just promise!

_33(Posted 2007) [#32]
Did you all know that SAM on the ATARI and Commodore 64 was a speech synthesizer that was taking, oh maybe 20K of ram, and it sounded JUST LIKE THIS ONE? :P

Are we totally inefficient or what?

John Blackledge(Posted 2007) [#33]
I know. I know.
Speak to MS.

AdrianT(Posted 2007) [#34]
Here's Paul and Daniel

Damien Sturdy(Posted 2007) [#35]

I am the greatest.

I knew it! You're DangerMouse, aren't you?

AdrianT(Posted 2007) [#36]
lol, now you mention it, that voice did sound vaguely familiar :)

puki(Posted 2007) [#37]
"Evak" hand that voice over - hand it over.

The first one is best - but I want both.

These ones:

Hand them over.