Cinematic-shot Skill: Plan Product UI Videos and Stills

A launch video with beautiful camera movement still fails to tell its story if viewers cannot understand what the product does. cinematic-shot starts with what happens in the UI before defining camera movement, lighting, and cuts.

A movie camera faces a plain paper stage, representing product cinematography

Cinematic-shot directs attention to real product behavior

Yiwei Ho’s cinematic-shot is an Agent Skill for directing product-launch visuals, especially app and website UI. The style uses close oblique angles, shallow depth of field, gradual camera moves toward events, and cuts following on-screen actions. The skill covers video and stills but includes neither a renderer nor a video-generation model.

This article cites 1weiho/cinematic-shot commit 7b9ec6c4 which contains visual direction guidance in references/look.md and shot-list examples separately. For similarly named skills such as cinematic-shot-direction, verify ownership because filming scope may differ.

The skill's direction can use a suitable rendering system. Actual video production still requires tools to capture UI, stage scenes, process images, and export reviewable files.

Identify what changes on screen before choosing camera movement and effects

The source starts with the sequence of events and the payoff: the moment viewers see what a feature does. For search, show both the query being entered and the relevant results. An empty search field or a rotating window does not demonstrate search.

State in one sentence: “Who will viewers see doing what, and what changes?” Choose verifiable demo content. For actual websites, show real capabilities without inventing buttons, states, or answers. Label hypothetical data as a demo and avoid implying business results.

Hypothetical brief: make a video of document search from the supplied demo page, showing query entry and results. Keep the site's actual light theme and specified demo data; invent no features/results. Send shot lists and sketches before full-quality rendering, plus a separate cover still.

Choose a stage that matches the product’s actual theme

The source recommends a dark stage for products with a dark theme and a light stage for products with a light theme. Preserve the actual UI before adjusting the lighting. If the website only has a light theme, retain it so buttons, borders, and text keep their original meaning.

The stage is the area around the UI plane and its lighting. Dark stages dim surroundings and illuminate focal points; light stages separate plane and background with restrained shadows and edges. Maintain consistent themes and lighting across shots unless changes have a reason.

Shallow depth of field directs attention to important lines or buttons while keeping them readable. Flat screenshot-edge blur is not simulated depth: focus must correspond to camera angle and planes.

A shot list makes the visual plan reviewable

The original workflow suggests a roughly 45–55 second video with 8–10 UI shots before text and the logo. This is the skill's direction, not a duration that guarantees campaign success. Users can adapt the brief by clearly stating their reasons and required deliverables.

ShotWhat you should seeCamera and focal pointCut point
Start a searchThe demo word actually appears in the search box.Close and angled, focused on the textAfter the key phrase is visible long enough to read
See resultsThe list changes according to actual searches.Move slowly toward the relevant itemAfter seeing the outcome status, not before knowing what happened
Open the documentThe selected item is directed to the correct content.Wider angle to see contextThe results should be understood before exiting the shot.

KrobClick presents this table as a hypothetical example, not shots actually created with the skill. Each shot should make clear its focus, event start, reveal, and reason for cutting. If viewers need lengthy explanations to understand a button, revise the story and demo data before adding camera movement.

Check installation and prepare the rendering workflow

The README provides these installation commands. Node.js and a supported agent are prerequisites. This article did not run the installer or render output. Capture and rendering tools have separate requirements to check for the actual production method.

npx skills add 1weiho/cinematic-shot

Close-up UI requires sufficiently high-resolution source captures or drawings. Stretching a small image across a frame blurs text and prevents feature verification. For animation, the source requires deterministic time-based frames for inspection and rerendering, rather than live events varying with network speed.

Before recording a page connected to a real account, prepare demo data or hide personal information. Distinguish viewing a page from actions that send messages, delete data, or have other external effects. Recording a video does not authorize those actions; check their consequences first.

Still images use composition instead of a moving camera.

The skill supports still images in the same style, but without timing and cuts, the composition and focal point must guide the eye. Set the target aspect ratio before choosing the framing. For example, a horizontal web image and a vertical image should not simply be cropped from one source without checking key positions.

The source recommends rendering still images separately instead of extracting video frames, which may contain motion blur or lack sufficient resolution. Choose a moment that clearly shows the action. Avoid ambiguous blur that looks like a rendering error, and keep visual effects from obscuring what the product does.

Check UI accuracy and clip clarity before the final render

Inspect a preview before the full render, both shot by shot and continuously: Is the focal point in the right place? Is important text readable at the intended viewing size? Can viewers follow the story across frames? Does the footage show actual product behavior or an explicitly labeled demo? The original workflow includes showing a preview and waiting for user approval before a full render, helping reduce revisions after substantial processing time has been spent.

Macro-shot counts and dark lighting do not show whether viewers understand a feature. If its purpose is unclear, revise the sequence before adding lights, lenses, or sound. Start with one working user path and make the shot list lead to that outcome.

Sources

Source reviewed October 21, 2026, pinned to cited commits. This explains documentation without installing or testing the skill. Thai examples are hypothetical KrobClick scenarios. The AI-generated cover is not output from the described skill.