Runway Gen-4.5 Text-to-Video: Shot Validation of Physics and Human Body
A capability illustration entry for the Runway Gen-4.5 text-to-video pipeline: a single shot is used to probe the three capabilities the model highlights in its release notes at once—strong prompt adherence, complex physics and causality, and realistic human motion. The shot is designed as one continuous camera move: a glass on a wooden floor is knocked over and shatters, then a person crouches down to pick up the pieces, all in 16:9 for 10 seconds. This scene was chosen because it combines rigid-body collision, flying liquid/debris, and human joint movement, making it easy for developers to observe cross-frame temporal consistency, objects appearing and disappearing, and whether the causal order is reversed (effect before cause)—exactly the known limitations of this model. On the input side, only the three fields prompt, aspect_ratio, and duration are needed, with no reference material, so it can be directly compared against the capability boundaries stated in "Model Facts"; it is suitable for rapid exploration of previsualization, storyboard validation, and shot controllability. Note that 9:16 vertical supports only 5 or 10 seconds; this entry uses the 16:9 plus 10-second combination to obtain a complete arc of action.
Style Text-to-Video · Cinematic · Physics simulation
Model Runway Gen 4.5
Prompt
Single continuous shot, 16:9, 10 seconds. A bright morning kitchen, wooden floor, soft window light entering from the left side of the frame. A glass of water sits near the edge of the counter; a person's forearm sweeps it off, the glass falls, hits the wooden floor and shatters, shards and water fly outward, with accurate gravity and momentum. The person then crouches into frame and carefully picks up the largest shard. The camera stays at a low tabletop height, slowly pushes in and tilts slightly downward to follow the falling glass, then settles steadily on the crouching person. Natural lighting, shallow depth of field, realistic skin and glass refraction, subtle handheld micro-movement, no cuts, coherent causal timing across all frames.
Copy the full prompt and adapt it for your own generation.
Tags