AI Video Production
0/24 complete

Module 04 · Generative Video With Veo

Shot Composition Prompts: Wide, Close-Up, Tracking

20 minfocused lesson7practical steps4grounded questions3source links
Open lesson + course map

On this lesson

Course outline

Shot composition tells Veo where the viewer stands, how much of the scene is visible, and whether the camera stays still or moves. “Cinematic” is not a composition instruction. “Wide shot, eye level, subject small in frame” is.

In this lesson, you will write six Veo prompts for one mini-scene, then arrange them as a storyboard. The sequence will use a wide shot to establish place, a close-up to reveal detail, and a tracking shot to add movement without losing continuity.

// concept

Build Every Prompt From Seven Parts

Write each shot as a compact production brief. Google’s current Veo prompt guide separates subject, action, scene, camera angle, camera movement, lighting, and other visual details. A practical syntax is:

// prompt — copy me4 lines
[SHOT SIZE + ANGLE] of [SUBJECT] [ACTION] in [ENVIRONMENT].
Camera: [ONE MOVEMENT OR STATIC]. Lighting: [SOURCE + QUALITY].
Look: [STYLE + PALETTE]. Duration: [LENGTH].
Continuity: [REPEAT THE LOCKED DETAILS].

Describe visible choices: “soft dawn light from frame left” is actionable; “beautiful lighting” is not. Keep subject descriptions unchanged across prompts. If your Veo surface currently supports reference images, use only an image you have permission to use and confirm its plan requirements in the official guide.

// concept

Choose the Shot for Its Editing Job

Composition is not decoration. Each framing choice answers a different viewer question.

Prompt termWhat it asks Veo to showBest editing purpose
wide establishing shotSubject small enough to read the surrounding placeWhere are we?
medium shotSubject roughly waist-up with some environmentWhat is the person doing?
tight close-upFace, hands, or object fills most of the frameWhich detail matters?
eye-levelNeutral human-height viewpointNatural observation
static cameraFrame does not moveClarity and an easy edit point
slow dolly inCamera physically approaches the subjectGradual emphasis
track alongside / truck rightCamera travels parallel to actionFollow movement through space
pan rightCamera rotates right from one positionReveal something beside the subject

For a clean three-beat edit, use wide → close → move: geography, important detail, then motion. Expand that rhythm into six shots:

// prompt — copy me2 lines
PLACE → ACTION → DETAIL → MOVEMENT → RESULT → END FRAME
 wide    medium    close     track       OTS       close

// concept

Protect Continuity Between Separate Generations

Veo may treat separate prompts as separate worlds. Create a short continuity block and paste it unchanged into every prompt:

// prompt — copy me3 lines
Continuity lock: same adult tea vendor, teal shalwar kameez, plain brown apron;
same red tea cart with turquoise hand-painted panels; same brick courtyard at dawn;
warm natural light from frame left; realistic documentary look; no visible brand logos.

Preserve screen direction: left-to-right travel should not reverse accidentally in the next shot. Generate one shot per prompt; a wide that becomes a close-up, circles the subject, and ends overhead contains competing moves.

Duration controls differ by Veo interface and model. Confirm the available settings. The sample’s six-second clips are a production choice, not an account-level promise.

// worked_example

Worked Example

This is a fictional sample for a Lahore tea-cart micro-story. The cart, vendor, and output notes are not a real business or reported generation result.

The weak draft prompt was:

// prompt — copy me1 line
Make a cinematic video of a Pakistani tea seller with dramatic camera movement.

It does not decide whether place or person dominates, while “dramatic” leaves the camera path undefined. The fix gives each shot one path and repeats continuity details.

// prompt — copy me32 lines
SHOT 1 — WIDE
Wide establishing shot, eye level: an adult tea vendor in a teal shalwar kameez and
brown apron opens a red cart with turquoise hand-painted panels in a Lahore brick
courtyard. Subject small in centre third; cart and courtyard readable. Static camera.
Soft dawn light from frame left, warm documentary look. Duration: 6 seconds. No logos.

SHOT 2 — MEDIUM
Medium eye-level shot: same vendor and clothing beside the same cart, placing a steel
kettle on a small stove. Waist-up; hands and kettle unobstructed. Static camera. Same
courtyard, dawn light from frame left, warm documentary look. Duration: 6 seconds.

SHOT 3 — CLOSE-UP
Tight close-up: same vendor's hands pour steaming doodh patti from the steel kettle
into a plain white cup on the red cart. Cup and stream fill frame; focus locked on the
rim, background softly blurred. Static camera. Same dawn light and warm documentary
look. Duration: 6 seconds. No text or logos.

SHOT 4 — TRACKING
Eye-level medium-wide shot: same vendor carries the white cup on a tray, walking left
to right across the courtyard. Camera tracks alongside at walking speed, constant
distance, no orbit or zoom. Same clothing, cart, dawn light, and documentary look.
Duration: 6 seconds.

SHOT 5 — OVER THE SHOULDER
Over-the-shoulder shot: same vendor places the white cup on a wooden table. Brown apron
in left foreground; cup sharp in centre. Static camera. Same courtyard, dawn light from
frame left, and warm documentary look. Duration: 6 seconds. No faces, text, or logos.

SHOT 6 — FINAL CLOSE-UP
Tight close-up: same white cup on the table, steam rising; red and turquoise cart soft
in background. Very slow dolly in, no rotation. Same dawn light and documentary look.
Duration: 6 seconds. Hold the centred cup for the final second. No text or logos.

The matching mini storyboard is:

ShotViewer learnsCut logic
1. WideLahore courtyard and tea cartCut once the place is readable
2. MediumKettle preparationMatch the vendor and cart from shot 1
3. CloseTea is the key detailCut on the pour for visual energy
4. TrackThe drink travels to someoneMaintain left-to-right direction
5. Over-shoulderThe cup reaches the tableCut on the placement action
6. CloseCup becomes the end imageLeave a clean hold for title added in the editor

Sample draft-one diagnosis: suppose shot 4 circles the vendor and hides the tray. Replace “tracking shot” with “tracks alongside at walking speed, constant distance, no orbit or zoom,” require an unobstructed tray, and regenerate only shot 4.

// failure_cases

Failure Cases to Diagnose

6 cases to diagnose

  • The wide looks like a portrait

    “Cinematic vendor” overweights the person. Add wide establishing shot, subject small in frame, and name the environmental features that must remain visible.

  • The close-up shows the wrong detail

    “Close-up of tea making” is underspecified. Name the exact object, action, frame fill, and focus point: cup rim, tea stream, hands, or steam.

  • Tracking becomes an orbit

    State the path relative to the subject: parallel, left to right, constant distance, no orbit.

  • The subject changes between clips

    Repeat clothing, age-neutral identity description, hero objects, palette, location, and light direction verbatim. Where available, test a rights-cleared reference image under the current official workflow.

  • The camera move is physically overloaded

    A track, tilt, zoom, and rack focus in one short shot compete. Keep one primary movement and move the other idea to another shot.

  • Cuts reverse direction

    A subject walking left-to-right followed by right-to-left can appear to turn back. Lock screen direction in prompts and note it on the storyboard.

// pakistan_angle

Pakistan Angle

For a Pakistani client shoot, agree before generation whether clothing, signage, food handling, and location should feel specific to Karachi, Lahore, Islamabad, Peshawar, Quetta, or another place. “Pakistani market” is too broad. Avoid generating fake Urdu shop text: keep signs blank and add approved Urdu or Roman-Urdu typography later in CapCut. Never put a real shop name, phone number, CNIC, vehicle plate, or customer face into a prompt without written permission.

Veo access, quotas, and USD-priced plans can change, so check the current official page before promising a client a generation count or turnaround. On unstable connections or during load-shedding, save the six prompts and storyboard locally, generate low-risk tests first, and download accepted clips as they finish. Queue final generations during a stable-power window; a mobile hotspot and repeated failed uploads can consume both data and paid quota.

// hands_on

Hands-On Exercise

7 steps

Build your own six-shot prompt list and mini storyboard.

  1. Choose one simple action without private client material.

  2. Write a continuity lock with subject, wardrobe/object, environment, light direction, palette, and screen direction.

  3. Plan six beats: place, action, detail, movement, result, end frame.

  4. Write one prompt per beat with all seven prompt parts.

  5. Include at least one establishing wide, one tight close-up, and one alongside tracking shot.

  6. Put the six rows into a storyboard with “viewer learns” and “cut logic” columns.

  7. Test one wide and tracking prompt, then record the defect and exact prompt fix. Without Veo access, use a peer review against the table. “Done” means the six prompts can be generated separately and cut in order without guessing the subject, camera path, screen direction, or purpose of any shot.

// completion_rubric

Completion Rubric

5 checks — tick as you verify

0/5

// sources

Sources

// check_yourself

Check yourself

4 questions · answers and options are taken word-for-word from this course

0/4
  1. 1 / 4 · diagnose

    Your work shows this failure mode: “The wide looks like a portrait.” The lesson describes it like this: ““Cinematic vendor” overweights the person.” What does the lesson tell you to do about it?