The 8-line prompt template (copy/paste) for Veo3Gen

A reliable way to reduce “randomness” is to structure the prompt so the model doesn’t have to guess what matters. This aligns with general prompt-engineering guidance: be specific, add relevant context, and keep language clear. (https://www.pcmag.com/explainers/prompt-engineering-101-the-secret-formula-for-writing-ai-prompts-that-actually)

Below is a canonical 8-field template you can paste into Veo3Gen and iterate line-by-line.

Subject: [WHO/WHAT is on screen + key identifiers]
Action: [ONE primary action + timing/tempo]
Scene: [WHERE + time of day + key props]
Camera: [shot size + lens feel + movement]
Lighting: [source + direction + mood]
Style: [realism level + references (broad) + texture]
Audio: [dialogue/VO + ambience + music + mix priority]
Constraints: [duration, aspect ratio, “no X”, continuity rules]

This mirrors common best-practice prompt structures that combine subject, action, scene plus optional camera, lighting, style. (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos)

Camera movement phrases you can reuse (and when)

Camera lines fail when they’re vague. Use one movement per shot unless you’re intentionally choreographing a sequence.

Locked-off tripod — use when you need stability (product, dialogue, typography).
Dolly-in — use to increase intensity or reveal detail.
Pan left/right — use to follow lateral action.
Handheld — use for documentary energy (expect some shake).
Orbit 180° / 360° — use to showcase a subject from multiple angles.
Macro rack focus — use for close-up detail and a deliberate focus pull.

FlexClip notes camera movement can include shot type, angle, and movement, and that multiple movements can be combined in one prompt—use that sparingly to avoid “camera soup.” (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos)

FAQ: Why does the video ignore my subject or change it mid-clip?

Symptom

Your character’s outfit, face, or key object details drift, or the subject is replaced.

Likely cause

Your Subject line is under-specified, or you’re accidentally giving competing descriptions across lines.

Exact line(s) to edit

Subject: add stable identifiers (age range, clothing, distinctive features, object materials).
Constraints: add continuity rules like “same outfit throughout” and “same subject, no replacements.”

Example rewrites

Before

Subject: a woman

After (Rewrite 1)

Subject: a 30s woman with short black bob hair, red raincoat, small silver nose ring, holding a yellow umbrella
Constraints: same person and outfit throughout the clip; no additional characters

After (Rewrite 2)

Subject: a worn copper teapot with a dented spout and a black bakelite handle, centered in frame
Constraints: keep the teapot design identical across all frames; no extra objects entering frame

FAQ: Why is there no motion (or the motion is random)?

Symptom

The clip feels like a still image, or motion happens in weird places (background swirls, unnecessary gestures).

Likely cause

Your Action line is too abstract (“walking,” “moving”), or the scene lacks physical cues the model can animate.

Exact line(s) to edit

Action: specify what moves and how (speed, direction, start/end beat).
Scene: add interaction props that “justify” motion (wind, traffic, tools).
Camera: choose a movement that reinforces the action.

Example rewrites

Rewrite 1 (add measurable action)

Action: she steps forward 3 paces, opens the umbrella, then pauses and looks up
Camera: medium shot, slow dolly-in

Rewrite 2 (constrain randomness)

Action: the teapot emits a thin stream of steam that increases gradually for 5 seconds
Constraints: no morphing; steam only from the spout; background remains still

FAQ: How do I control the camera without getting “camera soup”?

Symptom

The camera jitters, cuts unexpectedly, or combines several moves at once.

Likely cause

Your Camera line includes multiple conflicting instructions (e.g., “handheld drone shot orbit dolly zoom”), or your prompt describes multiple scenes.

Exact line(s) to edit

Camera: pick one shot size + one move.
Scene: keep a single location.
Constraints: ban cuts if you want one continuous take.

Example rewrites

Before

Camera: handheld drone orbit, fast zoom, whip pan

After (Rewrite 1: stabilize)

Camera: locked-off tripod, medium-wide shot
Constraints: single continuous shot; no cuts

After (Rewrite 2: one intentional move)

Camera: close-up, macro rack focus from foreground raindrops to her eyes
Constraints: no additional camera moves

FAQ: How do I keep lighting and color consistent across variations?

Symptom

One version is warm and golden; the next is cool and flat, even with the “same” prompt.

Likely cause

Lighting/color is implied but not stated. Lighting descriptions strongly influence mood and perceived depth. (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos)

Exact line(s) to edit

Lighting: specify source, time-of-day, direction.
Constraints: “consistent white balance,” “no dramatic exposure shifts,” “keep palette.”

Example rewrites

Rewrite 1 (clear lighting)

Lighting: soft overcast daylight, diffused shadows, cool-neutral white balance
Constraints: maintain same lighting and color temperature throughout

Rewrite 2 (mood + palette)

Lighting: warm morning light from camera-left, gentle backlight rim on hair
Constraints: warm palette only; no neon colors

FAQ: How do I get a specific style without breaking realism?

Symptom

You ask for “cinematic” and get something overly stylized, or you ask for “anime” and the scene becomes incoherent.

Likely cause

Your Style line stacks multiple aesthetics (e.g., “photoreal + watercolor + claymation”), or it conflicts with the Scene details.

Exact line(s) to edit

Style: choose a primary look and set realism level.
Constraints: explicitly forbid style drift (“no cartoon shading”).

Example rewrites

Rewrite 1 (realistic cinematic)

Style: photoreal, natural skin texture, cinematic color grade (subtle), no surreal elements

Rewrite 2 (stylized but controlled)

Style: minimalist 2D animation, flat colors, clean outlines, limited shading
Constraints: keep proportions consistent; no sudden texture changes

FAQ: How do I prompt audio (dialogue, ambience, music) without a messy mix?

Symptom

Dialogue is unclear, music overwhelms ambience, or the audio doesn’t match the scene.

Likely cause

Audio is underspecified: you mention “add music” but don’t describe priorities, tone, or what should stay subtle.

Powtoon describes Veo 3 as capable of native audio integration, including synchronized dialogue, ambient sounds, and background music. (https://www.powtoon.com/blog/veo-3-video-prompt-examples/)

Exact line(s) to edit

Audio: separate dialogue / ambience / music and specify mix priority.
Constraints: ban extra voices, specify “no lyrics,” or “music low under dialogue.”

Example rewrites

Rewrite 1 (dialogue-first mix)

Audio: Dialogue (primary): soft, close-mic voice saying “We’re almost there.” Ambience (low): light rain, distant traffic. Music (very low): slow pads, no lyrics

Rewrite 2 (no dialogue, ambience-led)

Audio: Ambience (primary): waves and gulls; Music (secondary): gentle instrumental, no vocals; no dialogue

FAQ: Text-to-video vs image-to-video: what lines change?

Text-to-video (most control)

A common 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)

Use the full 8-line template and treat Scene/Lighting as first-class citizens.

Image-to-video (protect the source image)

For image-to-video, FlexClip suggests a single-action structure: Subject + Action + Background + Background Movement + Camera Movement. (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos)

In Veo3Gen terms, you’ll typically:

Keep Subject short (describe what must not change)
Make Action minimal (micro-movements)
Add Background movement explicitly to avoid “everything animates”

Example (image-to-video rewrite using the 8 lines)

Subject: same person as the reference image; preserve face and outfit
Action: slow blink, subtle breathing, slight head turn to camera-right
Scene: keep original location from the image
Camera: locked-off tripod
Lighting: match the reference image lighting
Style: match the reference image style
Audio: room tone only
Constraints: preserve composition; background movement only: curtain gently sways, nothing else moves

A 10-minute “prompt debugging” checklist (what to edit first, second, third)

When a generation fails, don’t rewrite everything. Edit the most likely culprit line first.

Quick checklist

Subject: add 2–4 stable identifiers (clothing, material, unique feature)
Action: make it observable (start → change → end)
Camera: one shot + one move (or locked-off)
Lighting: specify time-of-day + source direction
Style: pick one look; set realism level
Audio: separate dialogue/ambience/music + mixing priority
Constraints: forbid cuts, morphing, extra characters, or palette drift

Do / don’t (to reduce conflicts)

Keep one location per clip
Use one primary camera move
Put “must-not-change” details in Constraints

Don’t

Stack multiple styles (“noir + anime + clay”) in one prompt
Demand multiple locations in a single continuous shot
List 10 camera moves and expect a clean result

Some guides note you can combine movements in one prompt, but it often increases chaos—start with one movement and add complexity only after you get a stable shot. (https://help.flexclip.com/en/articles/10326783-how-to-write-effective-text-prompts-to-generate-ai-videos)

Start generating via the API: /api
See plans and pricing: /pricing

Veo3Gen Prompt Framework FAQ (2026): 25 Questions to Fix Drift, Jitter, and “Wrong Style” Fast

Try Veo 3 & Veo 3 API for Free