What to make
The easiest way to make something people want is to make something you want. What do you wish existed that doesn’t? For example, back in the early 90s a friend of mine wrote some software for converting voice to data so he could talk to his girlfriend in Taiwan without paying for long-distance phone calls. That would have been a great project to turn into a startup.
If you’re a little older and/or have some particular domain expertise, you could try making something that you yourself don’t want, but you think other people do. This is a much riskier path, though. Most of the great startups seem to have begun with something the founders wanted: Google, Yahoo, Apple, even Microsoft.