Yeah, that’s actually a neat problem. I’d start by scraping sky‑condition data from weather APIs—like UV index, cloud cover, and even satellite imagery. Then feed the dates and times into a small neural net that scores each scene on “clean sky” versus “dramatic clouds.” You can even add a rule‑based layer for sunrise/sunset times so the lighting matches your montage’s mood. Once the script outputs a list of dates, just pull the corresponding footage or HDR sky tiles, and you’re set. Need a skeleton in Python or something? I can whip one up.