What comes a lot to my mind is what Stravinsky answered when he was asked about his composing process. If he knows before he starts what he wants. He answered, that he actually can not say that he knows what he wants, but he can definitely say what he does not want.
So while researching, I just try a lot of things and filter out what does not work for me. I also have the feeling that some compositions work better and others less with the type of visuals I like.
In august I did a project with the music of Hans Eisler. He was a pupil of Schönberg and presented him to his 60 birthday a composition called "Vierzehn Arten den Regen zu beschreiben". Which means 14 ways to describe the rain. It is for violin, piano, flute, cello, viola and clarinet. The same instrumentation like Pierrot lunaire. There was a visual artist that did a really nice video with water and rain (black and white). As the piece is only about 14 Minutes, but very dense, we played it 2 times: once without visuals and the 2nd time with. It worked really good. Unfortunately I think there is no video of the performance. But I will ask, maybe there is something.
J.S. Bach is also something that somehow works very good. Maybe because both - Bach and the twelve-tone technique of Eisler - have a strong mathematical, structural background? I don´t know.
This is something what comes out while I work on that topic. Still too rough, cheesy and monotonic. But the important thing for me is the researching/searching/finding/doubting process