3 réponses à ce sujet

[David Gaider]

TTS has become much better, and we use it in the initial stages of development so we can hear the lines being spoken out loud prior to recording them. The result is less robotic than it used to be... but it is still fairly monotone. Thus trying to insert TTS into a regular VO'd line is difficult unless the VO was recorded in the same monotone... and even then you'd have difficulty with the overall timing of the line, depending on how long the character's name was.

But could it be used in the future? Maybe. I've no crystal ball, but it seems like it might be feasible at some point. Hard to imagine it replacing real acting, though.

Modifié par David Gaider, 26 avril 2012 - 05:32 .

[Allan Schumacher]

Getting it to work without having the voice actor themselves say the name is where things are going to start falling down, because simulating a specific voice actor is where you're going to start hearing the irregularities and it will end up becoming jarring.

It is nice when games respond to your name, but the only ones I've seen it in are sports games, where it's the announcer recorded saying all sorts of different names (or if I share the same name with a different player). Trying to get the plethora of different voices in DA2 to properly say your name in the correct voice (and all that encompasses) isn't really all that possible yet, from what I can tell.

Modifié par Allan Schumacher, 26 avril 2012 - 07:05 .

[Allan Schumacher]

Might be more plausible to just have the voice actors pre-record a list of names(Like John, Mike, David, etc) and design a system that selects the name for the character in-game to use if it matches your character's name entry. Otherwise they just refer to you by the character's surname or a nickname.

This works really well with the sports games because those names are common. Heck, NBA 2k12 knows "Schumacher" because it's part of their valid names of German origin, which was pretty cool. (Schumacher for three.... drains it! \\m/)

Fantasy names allow people to have a ton of creative freedom with their name, which is going to trip up speech generators and we couldn't possible account for the infinite possibilities that players would create, so recording the individual names is inevitably provide limited coverage.

If/when we get to a point where TTS could be properly used to simulate an actor's voice, we'll probably be at a point where we're just using TTS for all the lines anyway haha.