Veo 3 Prompt "Shot-Language" Cheat Sheet: 25 Camera Moves + Framing Lines That Actually Change the Clip

TL;DR

If you want Veo to visibly change the clip, stop “adding more adjectives” and start adding one shot-language clause: a single, mechanical camera/framing instruction that’s concrete and non-conflicting. Below are 25 copy‑paste camera/framing lines grouped by outcome, plus a 3‑reroll A/B method to prove which phrases reliably stick (as of 2026-06-22).

Key takeaways

Veo’s own guidance explicitly asks you to specify how the output should be framed and how the camera should move during the shot—then experiment (https://deepmind.google/models/veo/prompt-guide/).
For consistency, change one camera variable per reroll (movement or framing or stability), while keeping subject/action/scene identical.
When a camera move gets ignored, it’s usually because the instruction is vague, stacked, or physically conflicting. Rewrite into one short clause with “only/no” constraints.
A practical prompt structure is Subject + Action + Scene + (Camera Movement + Lighting + Style) (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos).
If you’re making ads/UGC, readability constraints (center-safe, clean background, single subject) often outperform “cinematic” wording.

Why “shot language” beats longer prompts (and when it doesn’t)

Veo’s prompt guide doesn’t tell you to write essays. It asks two direct questions: How should it be framed? and How should the camera move? (https://deepmind.google/models/veo/prompt-guide/). That’s the core of shot language: one clause that communicates a single, filmable intent.

What shot-language is (in one sentence)

A short director-style clause describing either:

Framing (wide/medium/close, overhead, centered)
Motion (pan, push-in, orbit, dolly)
Stability (locked-off, stable horizon, minimal movement)

When longer prompts still win

Longer prompts help when you haven’t defined the fundamentals. The Gemini Enterprise Agent Platform guide says the most effective way to guide Veo is to break the idea down into key components (https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide). If your subject, action, or scene is mushy, a perfect camera move won’t rescue it.

Use this baseline structure (keep it boring on purpose):

Subject (the who/what the action revolves around) (https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide)
Action (one clear verb)
Scene (specific location details)
Camera (one shot-language clause)
Optional: Lighting and Style (Veo’s guide includes examples of both) (https://deepmind.google/models/veo/prompt-guide/)

Where to put camera directions (text vs workflow settings)

Put this in prompt text

Camera/framing belongs in the prompt text because it changes what Veo attempts to generate:

composition and framing
camera movement direction
stability/handheld vs locked-off

This matches Veo’s prompt-guide emphasis on framing and movement (https://deepmind.google/models/veo/prompt-guide/).

Put this in your generator’s settings

Use tool settings for delivery constraints like aspect ratio and resolution.

If you’re generating in Veo3Gen, you can choose:

Modes: Veo 3.1 Fast, Veo 3.1 Quality, Veo 3.1 Lite
Resolutions: 720p, 1080p, and 4K (4K on Fast/Quality)
Aspect ratios: 16:9 and 9:16
Inputs: text-to-video and image-to-video
Controls: first-and-last-frame control on Veo 3.1
Audio: native synchronized audio (dialogue/SFX/music) in a single pass
Pricing model: pay-as-you-go credits + optional monthly plans; purchased credits don’t expire; new users get free credits; developer API available

Mid-article CTA (natural + benefit-led): If you want to test these shot lines quickly, run the 3‑reroll micro-test in Veo3Gen using Veo 3.1 Lite for cheap previews, then switch to Fast/Quality when you’ve locked the motion.

The 25 copy-paste shot lines (grouped by goal)

How to use: paste one of these as your camera clause. Don’t stack moves until you’ve proven one works.

A) Reveal space (make the environment “read”)

“Wide establishing shot, slow pan left across the scene.”
“Start on a close detail, then tilt up to reveal the full subject.”
“Slow push-in from wide to medium, keeping the subject centered.”
“Overhead top-down shot, slow drift forward.”
“Crane up and back to reveal scale of the setting.”

B) Emphasize the subject (isolation + clarity)

“Medium close-up, subject centered, clean background.”
“Tight close-up on hands performing the action.”
“Side profile close-up, shallow depth of field.”
“Two-shot, both faces in frame, eye-level camera.”
“Follow shot from behind, keeping the subject mid-frame.”

C) Add energy (movement that feels intentional)

“Handheld feel, subtle micro-shake, walking pace.”
“Fast whip pan to the right, landing on the subject.”
“Orbit 180 degrees around the subject, steady speed.”
“Dolly sideways left-to-right, parallax visible.”
“POV shot, camera moves as if worn on the body.”

D) Reduce motion/warp (stability-first lines)

“Locked-off tripod shot, no camera movement.”
“Static wide shot, stable horizon, minimal motion.”
“Slow push-in only, no rotation or panning.”
“Center framing maintained throughout, no reframing.”
“Fixed focal length look, no zooming.”

E) Short-form ads (readability + center-safe)

“Center-safe composition, subject always centered.”
“Product hero shot, label facing camera, fills 40% of frame.”
“Clean seamless background, single subject, no clutter.”
“Waist-up framing, eyes on upper third, clear headroom.”
“Three-quarter angle product shot, slow turntable rotation.”

Veo’s guide explicitly recommends specifying framing/movement and then experimenting to see what feels right (https://deepmind.google/models/veo/prompt-guide/). Treat these lines as a starting phrasebook—and keep the ones that consistently change your clips.

Worked example: “meh paragraph” → controllable shot card (with A/B table)

Below is a concrete before/after you can copy and reroll.

Before (common failure mode)

“A woman introduces a new skincare serum in a bright bathroom, cinematic, high quality, smooth camera, nice lighting.”

What’s wrong:

“smooth camera” is not a move.
“cinematic” doesn’t define framing.
nothing forces label readability.

After (component prompt + single shot line)

Use the component approach recommended in Veo guidance (key components) (https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide) and common creator structures (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos).

Subject: Woman in her 20s, natural makeup, friendly tone.
Action: Holds a serum bottle, points to the label, then applies one drop to her cheek.
Scene: Bright modern bathroom, clean countertop, minimal clutter.
Camera: “Locked-off tripod shot, no camera movement.”
Framing: “Waist-up framing, eyes on upper third, clear headroom.”
Lighting: “Warm even lighting across the frame.” (lighting examples are discussed in Veo’s guide) (https://deepmind.google/models/veo/prompt-guide/)
Style (optional): “Slightly worn-out VHS tape texture.” (style examples are listed in Veo’s guide) (https://deepmind.google/models/veo/prompt-guide/)

Now test only the camera clause.

3-reroll A/B micro-test table (copy into your notes)

Reroll	Change only this line	What you’re judging
A	Locked-off tripod shot, no camera movement.	Is the label/face stable and readable?
B	Slow push-in only, no rotation or panning.	Does emphasis increase without drift/warp?
C	Handheld feel, subtle micro-shake, walking pace.	Does it feel UGC without losing readability?

If you’re doing this at scale, Veo3Gen’s developer API lets you generate variations programmatically while keeping your prompt versions tidy.

Image-to-video note (when you’re animating a still)

If you’re using image-to-video, a helpful single-action structure is: Subject + Action + Background + Background Movement + Camera Movement (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos).

EachLabs summarizes a similar idea for image-to-video: Subject, Action, Context/Environment, Cinematography/Camera (https://www.eachlabs.ai/blog/image-to-video-prompt-guide-best-practices-for-realistic-results).

Practical implication: when you already have a strong source image, keep the prompt focused on one action + one camera move.

When Veo “ignores the camera”: 7 rewrites that usually stick

Veo’s own guide encourages experimenting with framing and movement (https://deepmind.google/models/veo/prompt-guide/). These rewrites reduce ambiguity and conflicts.

Replace vague with mechanical
From: “cinematic camera” → To: “Slow push-in from wide to medium.”
Remove multi-move stacks
From: “pan and zoom and orbit” → To: “Dolly sideways left-to-right.”
Add “only/no” constraints
Use: “Slow push-in only, no rotation or panning.”
Anchor composition
Use: “Subject centered, framing maintained throughout.”
Simplify the scene when motion is failing
Use: “Clean seamless background, single subject, no clutter.”
Shorten the clause
If it reads like a paragraph, it’s not a clause.
Reorder into components
Subject → Action → Scene → Camera (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos)

Do-not-mix box (common conflicts)

Avoid pairing instructions that fight each other:

“Locked-off tripod” + “handheld micro-shake”
“Macro close-up” + “wide establishing shot” (in one shot)
“No camera movement” + “fast whip pan”
“Subject fills 40% of frame” + “tiny subject in vast landscape”
“Stable horizon” + “Dutch angle tilt”

Starter prompt blocks: 5 templates you can paste

These are intentionally minimal so you can isolate what changes.

1) Talking-head (9:16 creator intro)

Subject: [speaker description, outfit, voice vibe].
Action: Speaks to camera about [topic], one clear gesture.
Scene: Clean background, minimal clutter.
Camera: “Waist-up framing, eyes on upper third, clear headroom.”
Stability: “Locked-off tripod shot, no camera movement.”

2) Product demo (label readability)

Subject: [product], held by [hand description].
Action: Rotate to show label, then demonstrate one use.
Scene: Clean countertop, neutral background.
Camera: “Product hero shot, label facing camera, fills 40% of frame.”
Stability: “Static wide shot, stable horizon, minimal motion.”

3) Lifestyle b-roll (brand mood)

Subject: [person] using [product].
Action: [single action].
Scene: [specific location].
Camera: “Dolly sideways left-to-right, parallax visible.”

4) App UI moment (screen legibility)

Subject: Phone in hand, app open on screen.
Action: Tap one button, one screen change.
Scene: Clean desk, soft ambient light.
Camera: “Overhead top-down shot, slow drift forward.”
Constraint: “Center-safe composition, subject always centered.”

5) Narrative beat (character + location specificity)

Subject: [character with detailed appearance]. (More detail yields more specific results) (https://deepmind.google/models/veo/prompt-guide/)
Action: [clear action].
Scene: [thorough location description]. (location specificity is encouraged) (https://deepmind.google/models/veo/prompt-guide/)
Camera: “Crane up and back to reveal scale of the setting.”
Lighting: “Warm even lighting across the frame” or “Spotlight in one area of the shot.” (lighting examples) (https://deepmind.google/models/veo/prompt-guide/)

Checklist

Write Subject, Action, Scene as separate lines (key components) (https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide).
Add one shot-language clause (framing or movement).
If you need stability, add one stability clause (locked-off / stable horizon / no zoom).
For ads: add center-safe, clean background, single subject.
Run a 3‑reroll test changing only the camera clause.
Remove conflicts from the do-not-mix box.
Save winning lines into a reusable phrasebook.

FAQ

How do I get Veo to follow camera directions more reliably?

Use one mechanical clause (“Slow push-in only, no rotation or panning.”), and keep subject/action/scene stable so you can see whether the clause was honored (https://deepmind.google/models/veo/prompt-guide/).

How do I reduce warping and unwanted motion in AI video?

Prompt for stability explicitly (e.g., “Locked-off tripod shot, no camera movement” and “Stable horizon, minimal motion”) and avoid stacking multiple moves in a single shot.

How do I structure a Veo prompt so it’s not a messy paragraph?

Break it into key components—Subject, Action, Scene, Camera—as recommended (https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide). A common creator structure is Subject + Action + Scene + (Camera Movement + Lighting + Style) (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos).

How do I prompt camera moves for short-form ads without losing readability?

Use constraints that protect legibility: “Center-safe composition,” “clean seamless background,” and “product hero shot, label facing camera,” then keep motion minimal during the proof moment.

How do I test which camera phrases actually change the clip?

Run three rerolls, change only the camera clause (pan vs push-in vs locked-off), and judge one criterion (stability, readability, or energy). Veo’s guide encourages experimentation to find what feels right (https://deepmind.google/models/veo/prompt-guide/).

Are there safety or policy reasons a prompt might be blocked?

Yes. The Agent Platform guide states that Veo applies safety filters and that prompts violating responsible AI guidelines are blocked (https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide).

Closing CTA: build your “house camera language” faster

Pick 5 shot lines from the list above, run the 3‑reroll micro-test on a single concept, and keep only the lines that consistently produce the motion you want.

If you want an affordable way to access Google’s Veo 3.1 video models without enterprise pricing, generate those iterations in Veo3Gen—use Lite to preview cheaply, switch to Fast as your default, and reserve Quality for maximum fidelity. When you’re ready to scale variants, use Veo3Gen’s pay-as-you-go credits (purchased credits don’t expire), optional plans, and the developer API for programmatic generation.

Start creating with Veo3Gen

Veo3Gen gives you affordable Veo 3.1 video generation with native audio, up to 4K, and credits that never expire — with free credits to start.

Generate your first video now: Get started
Compare plans and pay-as-you-go pricing: See pricing

Veo 3 Prompt "Shot-Language" Cheat Sheet: 25 Camera Moves + Framing Lines That Actually Change the Clip

Try Veo 3 & Veo 3 API for Free