The Shot Test
What happens outside the workflows: field tests, observations, wrecks, and a few everyday uses that have nothing to do with productivity.
The Shot Test
Film Emulation for Video: Build a Look That Survives the Edit
Learn film emulation for video, from balancing footage and choosing a film look to adding grain and matching AI-generated shots in a finished edit.
The Shot Test
The Same Five Camera Moves Tested Across Four AI Video Tools
Runway Gen-4.5, Kling 3.0, Luma Ray3, and Sora 2 were tested on five fundamental camera moves—pan, push-in, tracking shot, tilt, and handheld—using identical prompts to evaluate which AI video model can execute specific cinematographic techniques rather than just generate visually appealing clips.
The Shot Test
Can AI Keep the Same Character's Face Across Six Shots?
ByteDance and Rice University's EntityBench 2026 benchmark tested 140 episodes across 2,491 shots and confirmed that AI video tools struggle to maintain the same character's face across multiple shots, with consistency degrading sharply beyond even a handful of cuts.
What Happens When You Ask AI for Natural Handheld Camera Movement?
AI video generators struggle to replicate the deliberate pause in handheld footage when the operator's body settles before reframing, testing Runway Gen-4.5, Kling 3.0, Luma Ray3, and Veo 3.1 across three documentary-style prompts to evaluate which model best understands human-operated camera motion.
The Shot Test
Rain, Smoke, and Reflections: Three Effects AI Still Struggles With
Runway Gen-4.5, Kling 3.0, Luma Ray3, and Veo 3.1 all failed at least one of three challenging AI video effects—rain, smoke, and reflections—revealing that each model understands physical phenomena differently despite impressive cherry-picked demos.
The Shot Test
How I Use AI for Previsualization Before a Real Shoot
A filmmaker explains using AI video tools not to replace production but to create complete previsualizations with actual camera movement and pacing before shooting, where approved frames become direct inputs to the pipeline rather than disposable reference material.
The Shot Test
Can AI Generate a Convincing Dialogue Scene Without Audio?
Runway Gen-4.5, Kling 3.0, Sora 2, and Veo 3.1 were tested on dialogue scene generation with audio disabled to evaluate whether AI models can produce convincing visual performances based solely on facial expressions, turn-taking cues, and listening behavior rather than relying on synchronized voice tracks.