text to speech synthesis

Blitz3D Forums/Blitz3D Beginners Area/text to speech synthesis

jhocking(Posted 2004) [#1]
Anyone know how to sythesize speech from text in Blitz? I'm not talking about particularly good speech mind you, just so that input text can be automatically spoken aloud. I think there are tools for this kind of thing floating around so I'm curious how to use them in Blitz. For example, is there a free text to speech library which can be compiled as a userlib?


puki(Posted 2004) [#2]
Didn't someone do this recently? - I seem to remember something.


jhocking(Posted 2004) [#3]
I don't recall anyone posting about text to speech, but now that you mention it I remember someone posting stuff for speech recognition. They were using Microsoft's speech SDK; I think that can be used for text to speech as well as speech recognition. Hm...


puki(Posted 2004) [#4]
Maybe that was what I saw - I know there was something about speech stuff recently.


jhocking(Posted 2004) [#5]
I've also found a tool called Mbrola. Can that be linked into Blitz?


puki(Posted 2004) [#6]
Maybe it can be done by outputting the text to a file (on-the-fly if necessary) and then calling a 3rd party speech program from within Blitz (but can you pass parameters to it? - perhaps if command line parameters are accepted by the 3rd party - but then will Blitz let you pass them out?)


jfk EO-11110(Posted 2004) [#7]
I wasted many houry in searching for a speech tool for blitz. I know there is a solution (guess a userlib from pepsi?) that is using the ms agent app that officially comes with windows since WinME. But infact it only works if that agent is installed correctly. It might work on all XPs, but not on all MEs and only on some Win98s, when the user manually installed the agent.

I was looking for a simple solution, some kind of exe file of some hundred kbytes. Didn't find anything yet.


_PJ_(Posted 2004) [#8]
2nd speech centre
------------ ignore - ------------------ didnt read post properly :)


Alberto(Posted 2004) [#9]
hello

It was me who posted about text to speech softwares some time ago, here and on GameDev.net.
People replied that syntetic human voices are not natural enough , even for a game application.
To me they sound reasonably good, but I am not an native english speaker, the Italian version was actually a disaster


jhocking(Posted 2004) [#10]
I'm not after natural sounding, I just need the read aloud text to be relatively easily understandable. Looong story short, this isn't for a game.


Alberto(Posted 2004) [#11]
no problem to understand it


electronin(Posted 2004) [#12]
when i look at the "speech" control panel, there is something that lets me set the default voice. maybe there is a way to access this thru the system dlls?


wedoe(Posted 2004) [#13]
Semar did some Blitz speech a long time ago but I never got his example over at Blitzcoder.com to work......


ckob(Posted 2004) [#14]
http://www.binary-people.com/app.php?id=8

beeps did this for some games and released it to the public check it out , it might help


jhocking(Posted 2004) [#15]
Thanks for the link but that's the speech recognition thing I mentioned above. I'm looking for text to speech synthesis, not recognition.


semar(Posted 2004) [#16]
@ Joe,
if you follow the link in my sig, there's something that may interest you, it's called So To Speak.

Sergio.


Physt(Posted 2004) [#17]
Check out my ChipTalk program at www.speechchips.com

It is an emulator for the old SP0256-AL2 chips. I could do a DLL version of that if there was enough interest.


Rook Zimbabwe(Posted 2004) [#18]
The MS Speech would not work if it was not installed * and activated * on someone elses computer.


Rook Zimbabwe(Posted 2004) [#19]
Back when I was teaching school all we had were older MACs and I hated them... You could write a text file though and get the computer to read it in those little annoying voices...

Then we realized that the voices increased and decreased their pitch in a set pattern -- like a song -- and we wrote lyrics that followed that pattern.

It is scary when a roomful of MACs sing at you!!!


jfk EO-11110(Posted 2004) [#20]
Physt - yes, I am highly interested in such a dll. ALthough it sounds way out of date, it has that "computer is talking" touch. It's kinda sexy :) reminds me of STSPEECH, or prof. Hawkins for some reason. A lot of hollywood movies use such synteticly sounding computer voices for speaking computers just to make sure the people will think it's "realsitic", tho these days it's possible to simulate a human voive pretty perfectly.

Yes, let me know when there's a DLL around, thanks. I needed it for a Computer Instructor voice.


cbmeeks(Posted 2004) [#21]
Buy and Amiga...lol

cb


jfk EO-11110(Posted 2004) [#22]
Amigas cannot be bought these days my friend, if they show up at all, they are exhumed.


Physt(Posted 2004) [#23]
ebay!


jfk EO-11110(Posted 2004) [#24]
touché :)


Sir Gak(Posted 2004) [#25]
I posted a request on blitzcoder.com not too long ago about this issue. I still have an Amiga, and had written a game that implemented the narrator.device so that text could be spoken. I was hoping somebody "out there" had created an engine for speaking text (it doesn't have to be natural sounding, just understandable). I don't think that referring people to the MS agent app will work for those who, like me, have WIN98. The MS speech app only works under later MS Windows, as I understand it, and not WIN 98.


jfk EO-11110(Posted 2004) [#26]
Yeah, that's right. They started with it in WinME, but my WinME doesn't support it, so it's not really supported on a wide range.

I started with a speech program some time ago, but it was pretty weak. It was based on syllables, like: mi me ma mo mu, ki ke ka ko ku etc. because it simply sounds bad when you use a seperate M and a seperate A for a single syllable like eg. "MA". Then again, the syllable "KA" could be assembled since there is a short pause between them: k-a.

My app also required to write things in a pre-pronounced way, using the german and latin way of spelling A,E,O,U and I. "ai am nad shur abaat tis".

Since it required all possible syllables (bi ci di fi gi hi li mi ni pi ri si wi shi ...) the total amount of wav-files that had to be loaded was more than hundred files. It became very slow. I dropped the project because A the output was not very good and B the filesize and loading time was unbearable.


LAB[au](Posted 2004) [#27]
Just bumped on this thread right now. Physt, I really like your chip emulator's voice. Let me know as well if there is a DLL available :)


Beaker(Posted 2004) [#28]
I would be interested as well.


AntonyWells(Posted 2004) [#29]
What gets me is out there, there is a text to speech app that sounds incredible.

So good I even used it for voice overs for a little 2d shooter..woman voice, reverb etc..and you could create .wavs.

but i can't find it or remember the name... But don't give up..it's out there somewhere.


Eikon(Posted 2004) [#30]
The easiest way is to use a vb script under XP:
http://www.blitzcoder.com/cgi-bin/ubb-cgi/postdisplay.cgi?forum=Forum10&topic=000747

*edit: Just read more of the topic and found this had already been vaguely mentioned. I know of no way on 98.


Perturbatio(Posted 2004) [#31]
Found a whole host of links here that may or may not be useful.


Physt(Posted 2004) [#32]
Still haven't had time to work on the DLL but I will. Been having far too much fun working on my single chip video generator....

I'm going to clean up the wave files and then have to find a replacement for BASS for audio output. My ChipTalk program is free and includes BASS. I don't want to require BASS for my DLL to work.

Actually, would you want Blitz to handle the audio playback. Maybe you would just call out to the DLL for text to speech and then a routine in blitz would take care of playing the wave file. I'm not sure if an external dll plays audio that it is going to mix correctly with Blitz.

Question to the people who are interested - Do you want a text to speech solution for blitz or just a simple dll for windows use? I don't think they would be the same thing.


Beaker(Posted 2004) [#33]
I think the point is to have a small package that does text to speech in blitz. Whether the sound is played in Blitz or not doesn't really matter.


LAB[au](Posted 2004) [#34]
Physt> Text to speech for blitz. I am using the BASS dll in blitz as well.