I was just reading a news article that says that Opera now includes voice interaction for it's Windows product. I never thought about that until now, but it sounds like a really great, useful feature.

I couldnt see anything indicating when it will be available to Linux, but I was wondering how voice support works within Linux. Does an application need to have voice support, or can you use an underlying program for a sort of translation layer?

Voice support is just another interface, simple commands will work without any changes, like "open file", "save", "print", "yes", "cancel" and so on.
There is no different for the application if you press a key or say a word, but your speech recognition program must handle these things.

It's like to add a mouse to a command line interface.

It will take a long time, till it will be ready, due recognition of simple words or clear definited sentences is easy, but many many word are so similar to others, that a time consuming training is necessary.
(Did you say "from cd-rom" or "format C" ? ;-)

Opera hasn't released the Linux version of that yet,... I don't think.

The interface is based on IBM's Via Voice technology, which is supposedly quite good (doesn't need to be trained with the user's voice for simple use). Personally, I'd like to see IBM GPL ViaVoice technology so that it could be intergrated as a part of the interface for all of Linux... But I dream...