I’ve been working through the various scenes and shots and rendering first drafts (in some cases with 10+ takes). I now have enough rendered to edit together a rough cut for “cold open” of the film, and you can watch it here:
It opens three hundred years before the main story, inside the Purple Tube on the day it died. It never explains itself. It’s an ordinary amber evening in a settlement where nobody is eating, because everybody is packing, and there’s a queue at a brand-new Gateway lock. The idea is that the audience comes out of it knowing more than Pelsa will for the rest of the film.
Most of these shots are first renders: Seedance 2.0 Mini at 480p, mostly built on simple Blender blockouts that set the camera, layout and motion, with a background plate and character references supplying the look. A few takes have been through the upscale pipeline (Real-CUGAN, with RIFE frame interpolation), and the rest are still at draft resolution. Some shots took one try. Shot 2.a took eleven.
What I’m looking at in this cut:
- Does it read without any explanation? The cold open has no exposition, so the images have to carry it.
- Pacing. The queue needs to feel long, and the waiting longer.
- Continuity across takes. Character consistency is still the hardest part of working with image-to-video models, and it shows in places.
- Gravity Not surprisingly, these video models really don’t know how to deal with zero gravity.
I’m going to continue through the rest of the film like this until I have a full rough cut - with all SFX and music - and then start the final post-production.