 | I'm a filmmaker in the traditional sense who's pivoted into using AI in my workflows and over time, I realised one of the biggest challenges I faced was creating consistent shot sizes and framing within a scene. So I spent the last few weeks reverse engineering my prompts based on actual Hollywood lenses and film stock (ARRI, Panavision, Split-Diopter, IMAX) to match how specific cameras and lenses are used across different shots, and also figured out how to generate still images with hints of movements on them that make it much easier to generate the videos more accurately. I've created an entire repository with a module of 50 cinematography references suitable for Midjourney which is now online. Using this approach has saved me quite a bit on token costs so I'm happy to share a couple of the prompts used in these reference images, and the entire database if you find them useful. I'll respond directly in the comments. Cheers! [link] [comments] |
