[LAU] Wanted: ideas for spatial, speech synth setup

Hi all, I want to setup a machine that "talks" (text-to-speech) with a number (say, 4 at a time) of parallel voices. The voices will be positioned dynamically (in between utterances) and freely in a room. As speech synthesizers, I think I will go for either rsynth or mbrola (rsynth, I think, eats more CPU and sounds less natural). Two setups come to mind: (1) I could try and use a number of soundcards and hard-wire each to an individual loudspeaker. (2) Or, use a single surround soundcard with its front and rear speakers, and a mixer. In setup (1), some shell script would decide where to place the voice and route the speech synthesizer output to the right soundcard. In setup (2), some realtime mixing (amixer...) of sounds would have to happen; the mixer's input pan would send the sound to some position in the room. Setup (1) sounds cool to me, since the hardware (soundcards) is inexpensive, and since each voice will come from a visible source. Have you made experiences with setups like these? Will a [rd]ecent PC cope anyway? Is a RT kernel needed? I've been talking about the software I know, but maybe some other program out there would work better? Thanks for your comments! Johannes -- gal.lery, lipsia, tyskland