Radiant Voice Acting : V - Skyrim

» Wed Sep 01, 2010 10:04 am

I wouldn't be that surprised if Bethesda included a cutting edge speech synthesiser package in the game, with a tweaked vocabulary/sentence set to smooth out the intonation and add inflection. The latest business versions of this sort of software are right at the cusp of passing for human and are actually small in terms of memory - minuscule in fact when compared with thousands of lines of pre-recorded voice acting.

Even if it doesn't appear in Skyrim, the next TES game will almost certainly include it. When each 'voice' only takes up a handful of Mb and can be further adjusted via variables to give a different tone, speed and so on, you can literally have a cast of hundreds for a handful of memory. Plus you can pump an near infinite set of procedurally created sentences through it.

Of course they'll still use big-name (or should that be big-voice) actors to create the pre-generated main quest speeches, but even if its not in Skyrim the future for radiant story will ultimately be speech synthesis.

Edit: Actually come to think about it, you could get away with some slightly weird (not quite perfect) intonation by explaining it as racial accents, or difficulties in pronouncing a 'foreign' language by non-nords.

As much as I would like to see this, it's not going to happen. As advanced as voice synthesis has become, you just can't get the same style and personality that you get with a voice actor. It would end up sounding much too generic and media sources would recognize this and give them lower ratings. They'd be better off sticking with plain text.

» Thu Sep 02, 2010 12:40 am

errr... wow, do you have a link to a demonstration of this? If that exists that would be the best thing ever... but... really? It exists??

Try the 'enter your own text' demo on the following website or search around for another... there are plenty of others out there. Its not quite perfect, but considering what its doing, its a remarkable piece of software. Taking the impressive rate of progression over the last couple of years, by the time the next TES game comes out it'll be near indistinguishable from a real human.

http://www.expressivo.com/

The current text to voice systems are only 2-3Mb for the most basic (which have little intonation or context pronunciation) all the way up to several hundred for a near lifelike, top quality one. Bear in mind too that you can drastically cut down on the phonetic database if you only have a limited pre-defined script vocabulary to intonate.

Assuming that after Skyrim players will expect voice acting for every conversation and demand dozens more different voiced actors to provide it, the only way forwards will be artificial synthesis - which will grant a memory, cost and versatility advantage over pre-recorded speech. Of course for very dramatic speeches you might want to keep a few professional voice actors to lend better gravitas, but for street conversation or shop keepers synthesis would be the way to go. Its just not practical to record (in multiple voices) that much dialogue otherwise.

» Wed Sep 01, 2010 7:11 pm

"Help me, my daughter has been taken by bandits. They have been known to use the caves in this area"

Indeed.

Let me mark it on your map!

» Wed Sep 01, 2010 1:55 pm

Try the 'enter your own text' demo on the following website or search around for another... there are plenty of others out there. Its not quite perfect, but considering what its doing, its a remarkable piece of software. Considering the impressive rate of progression over the last couple of years, by the time the next TES game comes out it'll be near indistinguishable from a real human.

http://www.expressivo.com/

The current text to voice systems are only 2-3Mb for the most basic (which have little intonation or context pronunciation) all the way up to several hundred for a near lifelike, top quality one. Bear in mind too that you can drastically cut down on the phonetic database if you only have a limited pre-defined script vocabulary to intonate.

Assuming that after Skyrim players will expect voice acting for every conversation and demand even more voice actors to provide it, the only way forwards will be artificial synthesis - which will grant a memory, cost and versatility advantage over pre-recorded speech. Of course for very dramatic speeches you might want to keep a few professional voice actors to lend better gravitas.

I'll admit it's come a long way, but not ready for Skyrim for sure... Maybe though it could be improved with some directing instead of just having to read blind. I noticed that there were odd pauses sometimes, but maybe a "director" could tell it not to pause etc. If you could teach the synthesizer not to be tripped up by certain lines, it could certainly lower the size of recording you need to do. But... we're not there yet, maybe someday.