Actually, I've never had any problems with a hardware synthesizer. In my case, a Dectalk USB in serial mode emulating a Dectalk Express. It has its quirks, but spelling words is not one of them. So, I suspect the problem was more with software speech than hardware speech. I think software speech is more sensitive to timing, whereas with the Dectalk, all that is needed is to dump text at a serial port and let the synthesizer decide what to do with it.
