Voice: text to speech?

How would I tie in a text to speech program?

I thought about adding a second sound card and using a cable to route the sound back into the other card. You would then have to select which sound source to use in SL.

Worse case would be to use two computers.

I want to know the answer to this question also.

Sorry I don't have any answer to give.

The mixer is an output system. I am guessing that by doing this you are adding the sound to your speakers and not to an input sound source. Some sound may leak threw to the microphone channel threw the echo canceler or threw feedback, but that would be unpredictable.



That product creates sound from SL text. (SL text chat --> speekers)
or it can do voice to text. (microphone ---> Text chat with very humorous results at times)

Right now I am thinking the only certain way to do it without software conflicts is to use an analog cable and multiple computers.

1. Scenario 1.
Typed text --> SL sound --> Internet --> other user
I imagine that software will be able to do this without extra hardware.

2. Scenario 2.
Human voice --> Phonetic text --> SL sound --> Internet --> other user.
This would probably need a second computer due to the speech to text conversion due to the processing time needed. Then again SL generally
does not use the second core anyway.

i have it and have used it some. it uses your built in windows text to speech system, so it will have whatever voices you already have installed (start-settings-control panel->speech to see and hear what voices you've already got installed). i only had 2 or 3 voices installed, and they were mostly bad. i downloaded a couple other ones that sounded a bit better. none of them are near "real sounding" though, if only because it's impossible for it to read tone and emotion from your text, so they all talk kinda like Spock.

that said, while none of them would pass to an observer as a real human, most of them ARE quite easy to understand, and it's kinda nice for keeping up with chat while tabbed out doing something in another program. it's also fun to stand around typing things and laughing at how funny they came out of the text-to-speech thinger. it's also cool you can assign different voices to different objects/avatars/UI notifications (assuming you have a few voices installed).

i haven't used the speech-to-text side of it, so can't comment on that part.

Technically, you should on most modern sound cards be able to use the mixer as an input source: on my Audigy 2 it's done by setting the input source to 'What U Hear' (What? Do you pay by the letter when you buy Creative or something? Where'd my 'yo' go?). I've done it for things like rendering MIDI tracks I've composed to Wave files, or playing a Wave file while I manipulated a VST or DirectX effect over it in realtime, recording the output.

It wouldn't solve your problem, though, since it's not selective: it would pick up all sounds going through the mixer, not just the TTS engine. So that would be ambient sounds in SL, any music you're listening to on Winamp, whatever else. So, you could turn off all sounds except the TTS, but that's hardly ideal.

Any software that could do it would have to trick the sound card on a very low level into thinking that the TTS engine was not internal. I don't know about low-level soundcard architecture, so I don't even know whether or not that's possible, but I would guess not.

I'm thinking two sound cards sounds like the best option so far, really, set the TTS to use the secondary card, then a line from one to the other, and set your input source as 'line in'. That should work, provided you don't get conflicts between the cards (conflicts? How very Windows 9.

Wouldn't it be easier to just not use voice if you're not going to speak into it? If you want to use text, then use text!

Text-to-speech would be nice if SL were to read out text you received on the other end, but having a poor imitation of a voice reading text and sending it as sound just isn't worth the bandwidth it wastes. Send the text instead!

yea i wouldn't send it as sound. if the person on the other end wants to hear all text as speech, they should get something like sl speech tools and run it themselves to translate it on their end without the bandwidth overhead.