Most of my scenes are heavily influenced by the directors I admire; Wong Kar-wai, Tarkovsky, Kurosawa, and others. That naturally shapes my creative choices and the perspective of each shot.
Here’s been my workflow so far: - Start with an idea and refine it into a script – Break the script into individual scenes, then into specific shots – For each shot, work with GPT to conceptualize a reference shot – Refine that until it feels right, then generate a MidJourney prompt – Iterate the prompt until the visual matches the emotion – Stitch everything together with an impactful soundtrack and simple editing – Viola!
If you’re deeply into cinema or anime, prompt engineering comes more naturally, it's less about keywords and more about emotional composition and visual storytelling.
That said, prompt engineering isn’t easy. There are tons of tweaks and hacks needed to get a decent shot. And even then, the biggest challenge is maintaining consistent characters and world design across shots. There just aren’t enough tools yet to give creatives full control over scenes.
I think there’s a real need for a tool that handles everything end-to-end, from script to visual storyboard to generation to editing/export.
It’s also a bit expensive. I blew through credits quickly while chasing the "perfect shot". Eventually, I had to settle for "good enough".
Overall, I’m incredibly bullish on AI generated films. We’re on the edge of a creative explosion, and the possibilities are endless.
And here’s a thought, if you believe AI won’t replace programmers because real dev work needs problem solving, then the same logic should apply to creatives. At its core, both are acts of structured creativity.
Would love to connect with anyone building in this space!