Text to voice for game dialogue?

[5ivexxxxx]

Well, stuff that can't be translated, or catchphrases, stay the same. So, yes, Rika still says Nipah. But really, if people were to watch everything in the "true" way, then most shows would make no sense to people. I recall in Metal Gear Solid 3, one character (whom I cannot remember for some dumb reason) commented on Snake's fluent Russian. Even though the dialogue was in English. The same goes for Gunslinger Girl. One minor character is mocked for being a poor Italian speaker. And that scene was hilarious, because I was watching it in Japanese, subtitled in English.

 

Also, did you watch the Deathnote dub? It's widely considered better than the sub. Some characters just work better. But people shouldn't let bad voice acting get in the way of their enjoyment of a game or show. The writing can still be good. For example, Resident Evil, a good game, has terrible writing and voice acting, whilst Elfen Lied, a bloody amazing show, just has bad voice acting (for the most part). But sometimes, voice acting is just sickening...

:sick:

[rdamstrom]

I guess its all just a matter of taste,while we would all like to have movie stars to do our voice acting the reality is that most mods will have none and I personally would have a robot sounding voice then none at all. this is a link to open Mary not as nice as Audrey who I really like but its opensource so there should be no problem distributing it. if nothing else it serves as a stepping stone to get the mod out and if it is well liked then it will be easier to get the voice actors to come to you. I will be using this in my mod to create an actual player / companion conversation as trying to say what I want in 43 characters just isn't working.after playing around with it while I discovered that they can say individual words much better then sentences so a little cut and paste may help

http://mary.dfki.de/

[gsmanners]

If push comes to shove, you can use speech synth to make your lip files and then just leave silence in the actual ogg file (and then say in the readme that the NPCs are "whispering").

[roquefort]

I guess its all just a matter of taste,while we would all like to have movie stars to do our voice acting the reality is that most mods will have none and I personally would have a robot sounding voice then none at all. this is a link to open Mary not as nice as Audrey who I really like but its opensource so there should be no problem distributing it. if nothing else it serves as a stepping stone to get the mod out and if it is well liked then it will be easier to get the voice actors to come to you. I will be using this in my mod to create an actual player / companion conversation as trying to say what I want in 43 characters just isn't working.after playing around with it while I discovered that they can say individual words much better then sentences so a little cut and paste may help

http://mary.dfki.de/

That Mary one isn't too bad actually. Still sounds a bit shaky on the diphthongs, but it might be possible to do some improvement on the result in Audacity, Wavepad or similar. Thanks for posting the link ;)

[Ihoe]

Speech synth'd voice is something I've tried, and it doesn't even make a believable robot. The best you can do is make a Stephen Hawking character and use it to voice that.

I disagree, I used one of these programs Unlicensed in one of my mods and people thought I have some professional voice acting assets and good voice actors, but All I used was synthetics.

Cepstral was the program I used, I must say, they were dumb to have it only say Buy "Licensed Cepstral..." in the middle of text to make sure it can't be reused, but audacity makes this vermin go away.

[gsmanners]

Wow. I really can't bring myself to use something like Cepstral without taking into account the way they intend it to be used. It seems like a disrespectful thing to sort of "hijack" it in that fashion. Surely, their developers deserve compensation for all the hard work they put into that program?

[roquefort]

Wow. I really can't bring myself to use something like Cepstral without taking into account the way they intend it to be used. It seems like a disrespectful thing to sort of "hijack" it in that fashion. Surely, their developers deserve compensation for all the hard work they put into that program?

Yes this might be drifting toward a violation of the `discussion of piracy' rule this site has. . . best change the subject :thumbsup:

 

If it's a demo version used for private amusement then I'd have thought it was OK though. That is the case for the ATT one.

[Ihoe]

It's not a Demo really, it only adds some additional advertising voices in the middle of your text, it's not a violation of copyright law anyway, besides, some guys like me really can't pay the full license for such things through the internet because we live in 3rd world countries that those sites don't support...

 

back on topic :D, I say if we have some voices, even if not professional or realistic, it's better than none at all.

[gsmanners]

I just tried out Cepstral, and you can definitely do a believable voice with it, although you need to learn SSML, tweak the cadence and maybe do some filtering here and there to fix the auto-tune sound you get in most of it.

[roquefort]

I just tried out Cepstral, and you can definitely do a believable voice with it, although you need to learn SSML, tweak the cadence and maybe do some filtering here and there to fix the auto-tune sound you get in most of it.

I tried it too; I hadn't heard of it before and it is impressive. I think you'd have to put full-stops after certain words to avoid odd-sounding transitions, plus further work as gsmanners suggests, but it could work. If you were going to use this and distribute the results legitimately it could get expensive though, since you pay per voice rather than pay once for the whole application. I suppose you could frequency-shift a voice to get some tonal variation.