A museum team has one approved illustration of a sculpture gallery and needs a five-second visual introduction. Should the camera stay locked while a visitor walks through, or should it dolly slowly toward the sculpture? Runway Gen-4.5 Image to Video uses a source image to establish the composition and a text prompt to define the motion. This guide compares two distinct camera strategies using a shot card, a single-motion prompt and a review grid. The gallery frames are illustrative drawings, not generated Runway output or evidence of a completed product test.
Choose one shot purpose before choosing motion
The opening shot must establish the sculpture’s position and the gallery’s clear exit path. If the narration explains the building layout, a locked camera can keep architectural relationships stable. If the narration emphasizes the sculpture’s material and form, a slow forward dolly might create emphasis. Write the visual anchor points: pedestal base, archway, floor tile lines and bright window. Select either an environmental action or deliberate camera movement as the primary change. Trying to demand a dramatic pan, orbit, zoom and character action within a few seconds adds ambiguity that a tightly controlled source image can often avoid.
Check the latest Gen-4.5 input and generation constraints
Runway’s Gen-4.5 documentation lists Image to Video and Text to Video, with image-and-text input for Image to Video. At the October 2026 review, it describes output duration choices from two to ten seconds, selected frame rates and supported aspect ratios; access depends on subscription. Those figures are a product documentation snapshot, not a warranty of a particular account’s remaining credits or render queue. For this scene, choose a source image that already fits the delivery frame rather than cropping the sculpture during generation. Save the source aspect ratio, target duration and frame-rate choices in the shot card.
Create a clean reference still without contradictory motion cues
Use an approved still depicting a stationary marble sculpture on a centered plinth, with coherent perspective lines and no motion blur. A crisp window reflection and visible floor guides help readers judge later geometry. Runway’s Image-to-Video Prompting Guide warns that unwanted motion cues in a reference can conflict with instructions. For a locked-camera test, avoid an image containing sweeping blurred lights or falling objects unless those are intended actions. Check rights, any depicted individuals and the source file’s crop. The supplied still defines visual content; the prompt should not spend its entire length redescribing the sculpture that is already visible.
Draft a locked-off camera prompt with minimal subject motion
Try a concise concept: “Continuous five-second shot. The camera remains locked on the gallery. Daylight shifts gently across the floor; sculpture and architecture stay still. No cuts.” This is a prompt example, not proof that any given generation will obey. The Runway guidance provides the phrase “The locked-off camera remains perfectly still” as one way to reduce unwanted camera motion, while acknowledging models may still move. Inspect whether fixed vertical edges stay vertical and the sculpture retains its shape. If the static result wanders or subtly zooms, record the failure and simplify rather than claiming true tripod-grade stability.
Create a separate slow-dolly version of the same shot
For a second concept, use the same reference still but change only the motion request: “Continuous five-second slow dolly forward toward the marble sculpture, keeping the pedestal centered; no cuts or orbit.” Runway’s camera terminology guide distinguishes dolly movement from a pan or tilt. In a dolly, foreground floor lines shift in perspective as the viewpoint approaches; in a locked shot they should not. Do not call a simple digital zoom a physically accurate dolly without inspecting parallax. Hold the still, duration and general scene constant so the editorial choice concerns motion rather than a different sculpture or background.
Compare the two concepts against the same checkpoints
Use four checks: identity of the sculpture, continuity of windows and pillars, camera path, and unwanted scene cuts. Inspect the first, middle and last frames, then watch in motion. In the locked version, drifting walls or a changing pedestal silhouette is a defect. In the dolly version, distorted sculpture arms, sudden lateral camera jumps or unmotivated objects entering the gallery are defects. Rank the result with written findings, not invented motion-accuracy percentages. A useful comparison sheet contains “intended motion,” “observed change,” “collateral change” and “editor verdict.” This makes the review reusable across later iterations.
Control unintended transitions and framing drift
If the image becomes a different gallery halfway through, check for language suggesting a new shot, fast motion or multiple actions. Runway’s official prompting guide recommends clarifying a continuous seamless shot and potentially increasing duration when unwanted cuts occur. Keep subject identity and architectural location stable before experimenting with complex moves. Do not try to solve a wrong crop by hiding the problem under a title overlay; return to the source still or movement specification. For delivery, protect space for a real headline in the edit rather than asking image generation to produce a reliable long text label inside the scene.
Handoff a transparent shot package
Save the authorized source still, exact prompt versions, model and generation settings, clips, selected frames, quality notes and final approved export. Confirm output resolution and aspect ratio before combining the shot with other video. State that the gallery motion is illustrative if there is any risk a viewer may mistake it for footage of a real location. Maintain the original image so a future editor can regenerate or replace the shot without guessing its composition. The meaningful outcome is an approved camera-motion decision supported by a documented check, not a claim that Gen-4.5 performed a physically perfect dolly every time.
How We Research
On October 28, 2026, we checked Runway’s Gen-4.5 model guide, its Image-to-Video Prompting Guide and camera movement terminology. We did not render or benchmark any gallery video. The sculpture, gallery and camera diagrams are original conceptual art; duration, output options and model access should be rechecked in the live account.
Frequently Asked Questions
Yes. In this mode the image establishes the scene and the text describes the intended motion.
No. Runway advises clear prompts but does not guarantee perfect stability.
No. A dolly moves the camera position while a pan rotates its viewing direction.
The referenced model guide lists durations from two through ten seconds.
Compare the same sculpture and architecture across frames, then watch for drift, cuts and geometry changes.