PARANOIA

Honestly, I considered this, then dismissed it as inconvenient. Plus I think it detracts from the impact of any surprise announcements if the players see you typing before each time Friend Computer joins the conversation.

OTOH, if you've already got the PC handy, you can use it to keep notes, and then they'll never know if you're typing notes or getting ready to blast them with an announcement.

Much better, IMO, if you were to prerecord some sound files of standard announcements and stick 'em, say, in an iPod hooked up to a speaker. Then you've got your Friend Computer announcements at the touch of a button.

Bear in mind that after a little thought, I decided against trying at all, so it's possible the impact of the synthesized voice overrides the disadvantages.
 

I've never used text-to-speech for conversations with The Computer, but I have used pre-canned recorded phrases (run through some audio processing) for public announcements and such - nice for ambiance, and not much at-table hassle (since I set it up to be "hit one key, service announcement X plays / is displayed").

I also used it for one mission briefing where the local CompNode crashed halfway through: after a brief freeze, the left audio channel started spewing incomprehensible low-level debugging information (hex digits and register data), while the right audio channel reported higher-level heuristics and response: "CompNode 91823 is down; attempting reboot. Reboot failed. Probability of Communist sabotage exceeds 18% - initiating Tier 3 Internal Security alert. Initializing hardware diagnostics. Hardware diagnostics failed. Secondary diagnostics offline. Probability of Communist sabotage exceeds 55% - mobilizing Armed Forces. Estimated time to containment of BNR sector: 2.8 minutes..." and so forth.

It was a lot of work, but so great to see the players go from an uncomfortable "Er, something's wrong with the prop" to "Wait, this is in-game" to "What the hell is going on" to "Armed Forces cordon? Kill on sight? I've got to get out of here!" Those taking too much time to listen to what was going on didn't make it out of the sector alive.

Doing that with text-to-speech would probably have been a lot less work. :)
 

I actually have used SayzMe in gaming sessions in the past with a great deal of success. First, I didn't tell the troubleshooters that I was doing it, so it had a pretty high hilarious factor the first time it came across.

It also helped me produce better games. Yeah, I know spontaneity is one of the high points in Paranoia, but it's still nice having a good framework. With that, I'll have dozens of phrases and such typed up well ahead of time to anticipate things. Including a number of canned responses. A quick ctrl+c/ctrl+v and it's talking. It does help being a fast typer too.

And when I would type something in, it didn't detract from the surprise. Instead, it actually elicited a stronger "aww crap" response, because they knew something was about to come, and it was going to be bad. Especially if I took my time.

You also have to know when to use it. As long as you and your players know it is a little bonus tool at your disposal, it won't hurt anything when you have to forgo its usage at the risk of slowing down a game. Usually my players get so involved with themselves and infighting, I have plenty of time to get stuff written to nuke them to all hell and back. But in a pinch, I'd just tuck my tongue to the roof of my mouth, and go all Stephen Hawking on them the old fashioned way.