I believe L_D's done a fair bit of work here. This is exactly the sort of work I do independently. Text to speech tools can be used with some effectiveness, but to attain the level being sought by most folks, you need a voice quality in the 64 to 128 bit level, and those packages typically cost around 50k for a six month lease of the service, as it requires multiple computers to effectively render it. They are also usually ignored in favor of custom voice work as the c-b ratio is pretty poor. Voice work is going to be your best solution given you are describing a specific want in terms of the way it sounds. You'll want a seriously well thought out script for each step, and you'll need an effective sound editor. Assemble the basic voice talent with a decent recording set up, and then basically record the dialogue to get the inflections, spoken at a slower than normal rate. Then, edit the dialogue itself into single word sound bites and reassemble as needed for the final result. Although you can use mp3 as a storage format if you needed to save space, for best results you'll want to save in 24 bit audio and use a lossless format. Typcial work for a single three path training video can take you up to 8 weeks to set up, and possibly 3 to render.

thou and I, my friend, can, in the most flunkey world, make, each of us, one non-flunkey, one hero, if we like: that will be two heroes to begin with. (Carlyle)