For me, it always starts with the words. Pictures may or may not come (I'm very rarely a visual writer, which means I have to concentrate very hard to get descriptions right - but also, I think means that I'm good at the small relevant details?). I'm much more likely to start with a clear first line and go from there, rather than a first image, if that makes sense? And then if the first line makes me ask questions about what's going on in the story, all the better.

For stories I'm not actually writing right now, it tends to be lots and lots of scenes with snippets of dialogue going on the whole time. Sometimes narration too, but mostly it's dialogue. Again, visuals are rare.