Working Elementally: Why Multi-Reference Is How I’m Making Films Right Now
Why I’m spending more time inside Seedance 2.0 Omni and Kling 3.0 Omni
I usually follow whatever workflow the technology rewards.
When the models rewarded timing and precision, I leaned hard into first-frame / last-frame workflows. I built animatics. I rendered scenes as stills. I animated between exact start and end frames because that was the best way to control the result.
That workflow still matters. For a lot of jobs, especially client work, it’s still one of the cleanest ways to get a predictable shot.
But the tools keep changing, so the workflow has to change with them.
Right now, most of my personal project time is going into the Omni models.
Seedance 2.0 Omni and Kling 3.0 Omni are leading the category for the kind of work I’m trying to do, and they’re changing the way I think about AI filmmaking.
Not just one image into one shot, but multiple references working together: character, environment, style, props, action, and sometimes previous frames from the film itself.
I’ve started calling this…
The Elemental Approach
The basic idea is simple: you build the elements before you generate the film.
Before I make a shot, I’m often building:
