Generative Video Shot Consistency
Plan text-to-video and image-to-video shots across providers, anchor recurring subjects with approved references, and verify visual continuity at each edit.
Start from the edit. Identify each shot's purpose, starting composition, subject action, camera movement and ending composition. For a recurring character or product, record a small reference set and a continuity record of invariants such as silhouette, outfit, markings, product geometry and palette. Use only references approved for the intended tool and use. Record the original file ID and the model or mode selected for the pilot. ## Key Points - [Runway Gen-4 video prompting](https://help.runwayml.com/hc/en-us/articles/39789879462419-Gen-4-Video-Prompting-Guide) - [Runway Gen-4.5 video modes](https://help.runwayml.com/hc/en-us/articles/46974685288467-Creating-with-Gen-4-5) - [Runway image references](https://help.runwayml.com/hc/en-us/articles/40042718905875-Creating-with-Gen-4-Image-References) - [Kling image-to-video API](https://kling.ai/document-api/apiReference/model/imageToVideo) - [Kling element management](https://kling.ai/document-api/apiReference/model/element)
skilldb get video-production-skills/generative-video-shot-consistencyFull skill: 34 linesGenerative Video Shot Consistency
Start from the edit. Identify each shot's purpose, starting composition, subject action, camera movement and ending composition. For a recurring character or product, record a small reference set and a continuity record of invariants such as silhouette, outfit, markings, product geometry and palette. Use only references approved for the intended tool and use. Record the original file ID and the model or mode selected for the pilot.
Match the mode to the shot
Text-to-video is useful for planning a look or producing a shot with loose visual continuity. Describe a concrete visible subject, one main action, camera position or motion, setting and lighting. Image-to-video starts from an approved composition in tools that use the still as a first frame. Concentrate the motion prompt on subject movement, camera movement and changes in the surroundings. If a tool accepts reference elements independently of the starting frame, use them for identity and set the first frame separately when the model and endpoint support that combination.
Runway's Gen-4 documentation requires a starting image. Its Gen-4.5 documentation describes text and image modes. Runway's image-reference workflow can help prepare a consistent still before video generation. Kling's API documents image-to-video, text-to-video and element management separately. Select an endpoint and supported element type for the current model. Keep capability notes specific to model and mode. Image counts, duration and native audio can change independently of the brand.
Pilot a continuity-critical shot
Pick a medium or wide shot that shows the defining features and one moderately difficult movement. For example, an original delivery robot has one amber circular light on its left flank and a short squared handle. In its pilot, it rolls a short distance right while the camera remains locked. The continuity record lists the flank light's side and the handle silhouette separately. Change one variable per retry so you can identify whether a failed candidate was caused by the reference, prompt or mode.
Inspect and assemble
Save the versioned prompt, approved input, job identifier and output ID together. Inspect first, middle and final frames, then watch the full motion at delivery size. Look for identity drift, shape swaps, lettering changes, duplicate objects, inconsistent direction and missing eyelines. Compare both sides of each planned cut. A generated end frame can seed another shot, but inspect it against the original reference before carrying its artifacts into a new generation.
Mark each take usable, fixable in edit or regenerate. For a subject mismatch, simplify the action, replace an ambiguous reference or cut to an approved insert. Use editing and compositing for exact graphics. Preserve approved original frames and log deviations that the project accepts. A playable clip, a visually consistent scene and a completed edit each need their own evidence.
Sources
Install this skill directly: skilldb add video-production-skills
Related Skills
Live Streaming
Techniques for producing professional live video streams, from technical setup and encoding to real-time audience engagement. Covers multi-camera workflows, redundancy planning, platform-specific considerations, and the discipline of broadcasting without an edit button.
Motion Graphics
Techniques for creating animated graphic elements for video, including titles, lower thirds, transitions, data visualizations, and explanatory animations. Covers design principles, animation timing, and the discipline of enhancing content without overwhelming it.
Veo Shot Prompting
Choose a Google Veo text, first-frame or subject-reference mode, write a testable short shot prompt, and verify the resulting video and audio.
Video Editing
Techniques for editing video content into compelling visual stories. Covers cutting, pacing, transitions, narrative structure in the timeline, and the editorial judgment that transforms raw footage into focused, emotionally engaging sequences.
Video Lighting
Techniques for lighting video productions to create mood, dimension, and focus. Covers three-point setups, practical and motivated lighting, color temperature management, and the craft of making artificial light feel natural while guiding the viewer's eye.
Video Preproduction
Techniques for planning video productions before cameras roll. Covers scripting, storyboarding, scheduling, location scouting, budgeting, and the organizational discipline that prevents expensive problems during production and post-production.