Geometric constraints make storyboard framing measurable, but action fidelity drops

Storyboard framing can now be measured and solved as geometric constraints rather than left to an image model's defaults: on 204 external director-storyboard shots, delivered head height can be brought to 0.955 times the target, and single-subject panels land within 1.2% of frame width of their declared positions.

Previously, screenplay-to-storyboard spatial planning relied on an image model's defaults, and delivered head height was 1.906 times the staged target from the director's words and 1.733 from the compiled prompt, leaving framing neither controllable nor measurable.

The PACE preprint uses a typed representation plus a camera solver to turn spatial planning into measurable geometric constraints; declaring the pose on 30 shots raised the action drawn from 58.9% to 74.4% without moving the framing. All of the above are the authors' first-party self-tests.

The authors note that the setting holding framing best draws the least described action, and transitions, fitted motion and human review remain open; v2 also removed supplementary material that was never uploaded, and the results await independent reproduction.