Communitygithub.com

gxgleo67/cinematic-skills

Master cinematic shot design like a pro—a knowledge base of camera choreography, dialogue scene blocking, directorial visual techniques, and analysis of classic scenes from Spielberg/Tarantino. Use this when designing shots, solving how to film dialogue/action/suspense scenes, or looking up shooting techniques.

What is cinematic-skills?

This skill provides a comprehensive knowledge base for cinematic shot design, covering camera choreography, dialogue scene blocking, and directorial visual techniques, including analyses of classic scenes from Spielberg and Tarantino. It is designed for AI agents assisting with filmmaking, video production, or visual storytelling, offering structured frameworks and decision-making guides. The skill is organized across 55 chapters divided into five volumes: low-budget cinematic tricks, dialogue scene mastery, director's vision, and dedicated sections for Spielberg and Tarantino styles. It features topic and chapter indexes, a cheatsheet for quick decisions, and pattern libraries, making it invaluable for solving specific shooting challenges such as action sequences, suspense, or power dynamics in dialogue. The skill supports parameterized queries (e.g., topic name or chapter number) to quickly access relevant content.

Works with~Claude Code~Codex CLI~Cursor
npx skills add https://github.com/gxgleo67/cinematic-skills/tree/HEAD/cinematic-shots

Ask in your favorite AI

Open a new chat with this agent skill pre-loaded.

Documentation

Cinematic Shots Like a Master

Content: 5 Volumes of Shot Techniques | Chapters: 55 | Generated: 2026-10-18

How to Use This Skill

  • No arguments — Load the core framework below.
  • With a topic — e.g., "chase scene", "dialogue", "suspense" — use the topic index to locate and read the chapter.
  • With a chapter — e.g., "vol1-ch03" — directly read the file with the same name under chapters/.
  • With a problem — First check cheatsheet.md (decision rules and branches), then patterns.md (operational details).

Core Framework and Mental Model

I. Solution Ladder: Blocking → Camera Position → Cutting

  • When a scene "doesn't look good": Try in this order. Cost increases at each level; the later the step, the more likely it exposes the director. (sp·General Principle; v1c6; v3c8)
  • When a single blocking can simultaneously convey relationship and emotion: Shoot it in one take. Symbolism (separating two people, power transfer) only works when you "don't cut." (v2c8)
  • When the performance is the entire value of the scene: Better to use one fewer camera position and let the actor complete the performance in one take. (v3c8)
  • Exception: Use multiple camera positions only when a "daily baseline" is needed for later disruption; coverage is capital, not a solution. (tq8)

II. Focal Length Before Movement

  • When you want the audience to see only one face and feel the world is pressuring them: Use a long lens from a distance, with other people pressed close to the lens as an out-of-focus foreground. (v3c1)
  • When you need to convey "can't run, can't escape": Long lens with a slight pan in the same direction to cancel out movement, positioning the camera so feet are not visible. (sp4)
  • When you need speed and chaos: Use a wide lens, allowing nearby objects to blur past as speed references. If the focal length is wrong, everything after is salvage work. (v1c2; v3c1)

III. Height and Distance: Two Power Coordinates

  • When two characters have unequal power: Place the camera at the height of the character you want the audience to identify with; both cameras at the same height. (v1c9; v3c9)
  • When you want to say "he is stronger" without dialogue: Make the closer person even closer (closer = power); let the in-focus person speak (focal plane = relationship). (v3c5)
  • When power or sympathy reverses during a scene: Have the character being looked down upon stand up and cross the frame, then unify the entire scene at the new height. Only do this once. (sp8)
  • When you need to convey "this space makes me uncomfortable": Only then allow height to not belong to anyone (height betrayal), once per scene. (v2c2)
  • When you want to bring characters closer together: First create distance, then let a single movement struggle to close it. When you want estrangement: Place a physical object between them that could be moved by hand. (v2c5; v2c1)

IV. Motivation and Movement

  • When you need the camera to move: First give an in-frame reason (someone stands up, turns on a light, crosses the frame); no motivation = the audience sees the camera. (v1c8; v3c11)
  • When you want to show an internal decision: Push in, cross the line, or reverse backward; the camera stops when they stop. Camera moving opposite to character = determination, camera moving back at same distance = inescapable. (v1c5)
  • When a push-in or orbit needs to carry emotion: Lock onto the coordinates of the eyes or face in the frame; stay on them for the entire shot. (v1c5)
  • When the entire scene is moving: First give stillness—a single movement after stillness is worth ten pans or tilts. (v1c5; sp2)
  • When there's no track but speed is needed: Move the camera in the opposite direction, using a static foreground as a reference. (sp10)

V. Information: How Much and When to Give It

  • When the audience must learn many things at once: First list "what must be seen, in what order," then use a single movement to bring them in sequentially. (v3c11)
  • When you want suspense: Let the audience know more than the characters (show the threat first). When you want a jump scare: Let the audience know no more than the characters (subjective shot + empty frame). (v1c4; tq3)
  • When you want to prolong a reveal: Show only part of the target object; let the character stop and look, and the camera stops too. When a section feels dragged: Delete the process, show only the reaction and the result. (sp4; sp9)
  • When key information lands: There must be a visible shift (height/focus/lighting/angle); otherwise, the audience won't know "something changed." (v2c6)

VI. Tension: Through Displacement and Rhythm, Not Fragmentation

  • When tension needs to hold for twenty minutes: Rank "invasions" by magnitude (sit down → pour drink → cross through → steal attention), one per step, getting progressively closer. (tq4)
  • When tension is already maxed out: Give an empty room or harmless result to release pressure, then immediately plant a new doubt. (sp6; tq7)
  • When you need to re-tighten: First return the camera position to the opening composition, then dismantle it—without a baseline, violence is just noise. (tq8)
  • When you want dread: First promise safety (ordinary composition + real actions that draw attention), then break the promise; the reason must come from the plot. Before the threat arrives, if an empty scene is held long enough, the audience will feel that "something is there." (v1c7; v1c4)

VII. Dialogue Scenes and Resource Management

  • When two people face each other at equal distance for shot-reverse-shot: Default to a different plan—two-shot, mirroring, blocking instead of reverse shots, simultaneous push-in, small-angle jump cut (see cheatsheet for options). (v2 entire volume)
  • When shot-reverse-shot is unavoidable: Push both cameras in at the same speed; pushing only one creates unilateral pressure. (v2c5)
  • When dialogue exceeds two minutes: Have the actors turn or walk; the turn should happen at a turning point and have motivation. When two characters shouldn't look directly at each other: Let them look off-screen or at an object they are holding. (sp5; v2c6)
  • When there are three or more people: You don't need to see every face clearly; just let the audience immediately know "who is the protagonist and who is in trouble." (tq10)
  • Close-ups and movement are scarce resources: Use a close-up only for the second when a character's inner thoughts must be read; if given to a strong character, it must also be given to a weak one. Reserve 1–2 movements per scene for turning points. (tq7; tq4)

Chapter Index

Volume 1: 100 Advanced Tips for Low-Budget Feature Films

FileKey Technique
ch01 Fight ScenesLong lens tricking
ch02 Chase ScenesEqual distance pressing
ch03 Character Introductions and ExitsReveal by blocking
ch04 Suspense Searches and StalkingWaiting in empty shot
ch05 Plot ChangesCamera-actor reverse
ch06 Appearances and DiscoveriesFocus handoff
ch07 Dread ShotsSafety violation
ch08 Guiding AttentionMovement motivation
ch09 Car ScenesSeat position = stance
ch10 Dialogue ScenesTwo-shot division
ch11 Arguments and ConflictsCross axis
ch12 Love and PassionDetails as substitute

Volume 2: 100 Advanced Tips for Shooting Great Dialogue Scenes

FileKey Technique
ch01 ConflictCross axis, obstacles
ch02 TensionEqual-distance circle shot
ch3 ConfrontationBoundary space
ch04 Group ConversationsRelationship shot
ch05 Building ConnectionSimultaneous push-in
ch06 Revealing ChangeFocus shift
ch07 Walking and TalkingCamera distance differential
ch08 Strong EmotionsSingle-shot symbolism
ch09 IntimacySmall-angle jump cut
ch10 Distant ConversationsRear observer
ch11 Creative ScenesInsert between two people

Volume 3: Director's Vision: 100 Camera Choreography Moves

FileKey Technique
ch01 Advanced Lens UseLong lens close/far
ch02 Basic MovementsActor-driven
ch03 Depth BlockingThree layers of depth
ch04 Advanced FramingCross axis, head/tail room
ch05 Symbolic BlockingDual push
ch06 Art and LocationsFake walls, reverse establishing
ch07 Dynamic ActionForeground sweep
ch08 Shooting PerformanceCamera economy
ch09 Camera HeightHead height
ch10 Complex MovementsOpposite sliding
ch11 Advanced TechniquesStory point sequence

Spielberg Section

FileKey Technique
ch01 Chase UrgencyBackground handoff
ch02 Guiding Eyes with ObjectsObject-driven pan
ch03 Conceptual InformationDual-purpose establishing shot
ch04 Visualizing DiscoveryHalf-reveal delay
ch05 Lively Dialogue BlockingMicro-scene
ch06 Layered SuspenseTension-release wave
ch07 Depth BlockingLayered crowd
ch08 Symbolic HeightPoint-of-view shift
ch09 ConcisenessReaction only
ch10 Visual StorytellingReverse movement

Tarantino Section

FileKey Technique
ch01 Building TensionFlat establishing shot
ch02 Subtle ConflictSide long take
ch03 Generating ExpectationSame-position jump cut
ch04 Suffocating TensionInvasion relay
ch05 Minimal CuttingMirror two-shot
ch06 Impending ViolenceVisualizing stakes
ch07 Intentional Anti-ClimaxSame composition interrupt
ch08 Breaking RoutineBaseline return
ch09 Manipulating SpaceMaster shot power dynamic
ch10 Group DialogueAttention handoff
ch11 Loss of ControlFake-out violence

Topic Index

  • Violence/Action → v1c1, v3c7, tq11
  • Dialogue Scenes → v1c10, v2c1–c9, sp5
  • Multiple Characters/Group Scenes → v2c4, v3c11, tq9
  • Focal Length → v1c1, v3c1, sp4
  • Height/Power → v1c9, v2c2, v3c9, sp8
  • Shock/Suspense → v1c4, v1c7, sp6, tq3
  • Entrance/Appearance → v1c3, v1c6, sp3
  • Focus/Shift Focus → v1c6, v3c5, tq5
  • Movement/Motivation → v1c5, v3c2, tq4
  • Space/Axis → v2c1, v3c4, tq9
  • Chase/Speed → v1c2, sp1, sp10
  • Blocking → sp5, v2c7, tq2

Supporting Files

Scope and Limitations

Covers only the refined reconstruction of the original book content (not verbatim excerpts). The original books are illustrated; overhead views and stills were not OCRed. For composition details, prompt the user to consult the original book. Some techniques depend on on-set conditions (dolly, crane, fake wall); adapt based on available equipment before application.

Abbreviation key: v1/v2/v3 = Volume 1/2/3, sp = Spielberg section, tq = Tarantino section, c7 = Chapter 7.

Individual skills in this repo

This repo contains 1 individual skill — each has its own dedicated page.

Related Skills