Prompts2026/05/20

Best Gemini Omni Flash Prompts for AI Video Generation

Explore Gemini Omni Flash prompt examples for cinematic AI videos, physics-aware scenes, consistent characters, conversational edits, product ads, and YouTube Shorts.

Best Gemini Omni Flash Prompts for AI Video Generation

Quick Answer

Good Gemini Omni Flash prompts aren't single sentences — they are short scripts. The prompts that get the best results spell out the scene, the camera, the lighting, the physics, the audio direction, and what to avoid. This post is a copy-ready library: cinematic clips, physics-aware scenes, consistent character shots, conversational edits, product ads, and YouTube Shorts.

Key Takeaways

  • The best prompts read like a shot list, not a tweet.
  • Always include camera movement, lighting, and a negative prompt — these three alone remove most of the "AI slop" look.
  • For multi-shot work, lock down the character description once and reuse it verbatim.
  • For physics scenes, describe the motion behavior, not just the object.
  • For Shorts, write 9:16 vertical explicitly and keep the duration short.

How to structure a great prompt

A reusable structure that works across cinematic clips, product ads, and character work:

Subject:   <what is in the shot>
Action:    <what is happening, with motion verbs>
Scene:     <where it is, time of day, environment>
Camera:    <shot type, movement, focal length feel>
Lighting:  <key light, mood, color temperature>
Style:     <cinematic / realistic / anime / product-ad / UGC>
Physics:   <gravity, friction, collision behavior, if relevant>
Audio:     <what the shot sounds like>
Aspect:    <9:16 / 16:9 / 1:1>
Duration:  <8s / 10s>
Negative:  <what to avoid>

You don't have to fill every field — but listing the dimensions in your head before you write the prompt is what separates "OK output" from "the shot you wanted".

Cinematic clip prompts

Cinematic realism: glass sphere on a ramp

A realistic 8-second vertical video of a glass sphere rolling down a wooden
ramp. It accelerates naturally, rotates as it moves, bounces once on a rubber
surface, then comes to a natural stop. Cinematic lighting with soft shadows,
shallow depth of field, slight camera parallax. 9:16, 8s.

Negative: floating objects, broken reflections, unrealistic gravity, jittery
motion, inconsistent shadows, distorted geometry, flickering textures.

Cinematic realism: rainy street at night

A cinematic 10-second tracking shot down a rain-soaked city street at night.
Neon signs reflect in the puddles. A lone figure walks away from camera, hands
in coat pockets. Shallow depth of field, anamorphic lens feel, warm tungsten
key light against cool blue ambient. 16:9, 10s.

Audio: rain on pavement, distant traffic, no music.
Negative: melted faces, floating shadows, mirror-letter signs, glitching neon.

Physics-aware scene prompts

Pouring liquid into a glass

A close-up 8-second shot of clear water being poured into a tall empty glass on
a marble countertop. The stream is continuous, hits the bottom, builds up
realistically, and forms small ripples and a meniscus at the surface. Soft
window light from the left. Cinematic style, shallow depth of field. 1:1, 8s.

Physics: liquid follows gravity, fills the glass progressively, no splashes
that defy momentum, no liquid passing through the glass.

Negative: floating water, sudden teleport-fills, broken reflections, ghosting.

Falling dominoes

A 10-second side-tracking shot of black dominoes falling in sequence on a clean
white surface. Each domino strikes the next, transferring momentum naturally.
Subtle camera parallax as the chain progresses. Studio lighting, hard shadows.
16:9, 10s.

Physics: realistic mass and collision, no early or skipped falls, no dominoes
floating between contacts.

Negative: floating pieces, motion-blur smearing, geometry distortion.

Consistent character prompts

For multi-shot consistency, lock down the character description as a reusable block and paste it verbatim into every prompt.

Character block

Character: Maya — a 28-year-old woman with shoulder-length wavy dark brown
hair, light brown eyes, warm olive skin, a small silver hoop in her left ear,
wearing a fitted navy denim jacket over a white t-shirt and dark jeans.
Confident posture, calm expression.

Shot 1 — coffee shop

[Character: <paste the block>]

An 8-second medium shot of Maya at a wooden coffee-shop table, taking a sip
from a ceramic mug, then looking off-camera and smiling. Warm morning light
from a window on the left. Shallow depth of field. 9:16, 8s.

Negative: face shape changes, eye color shifts, outfit changes, hair length
changes, jewelry inconsistency.

Shot 2 — city street

[Character: <paste the block>]

An 8-second walking shot of Maya on a sunlit city sidewalk, hands in jacket
pockets, walking toward camera at a relaxed pace. Light golden-hour glow.
Shallow depth of field, slow dolly-in. 9:16, 8s.

Negative: face shape changes, eye color shifts, outfit changes, hair length
changes, jewelry inconsistency.

The character block is identical. The model has the best chance of recognizing Maya as the same person when the input description does not drift between shots.

Conversational edit prompts

These prompts are written as instructions to a clip that already exists, not as new scenes. Use them with a workflow that supports edit-style prompts.

Edit: keep the existing scene, camera, lighting, and timing.
Change only: the character's jacket — from red to navy blue.
Preserve: face, hair, posture, outfit silhouette, and all background details.

Edit: keep the existing scene and character.
Change only: the camera — pull back 30% so we see the full upper body instead
of a close-up. Maintain the same focal-length feel.

Edit: keep the existing character and outfit.
Change only: the scene — move from coffee shop to a sunlit park bench. Keep
the time of day and the warm lighting.

Product ad prompts

Skincare bottle in water

A premium 10-second product shot of a luxury skincare bottle floating in
crystal-clear water. The bottle rotates slowly. Soft caustics ripple across
the label. Cinematic backlight from above, dark gradient background. Shallow
depth of field. 9:16, 10s.

Physics: realistic buoyancy, water surface tension, subtle bubbles trailing.
Audio: soft underwater ambience, no music.
Negative: warped label text, broken reflections, melting bottle geometry.

Sneaker on a turntable

An 8-second studio shot of a single white sneaker on a slow-rotating turntable
against a soft gray gradient. Hero product lighting, hard rim light, soft fill
from the front. The shoelaces and stitching are crisp. 1:1, 8s.

Negative: extra shoes, doubled tongues, blurry text on the heel, distorted
sole geometry.

YouTube Shorts prompts

Shorts thrive on one idea, one hook, one shot. Keep the prompt tight, prioritize 9:16, and write a strong negative prompt — the platform compresses the video aggressively, so any AI artifacts get magnified.

Visual hook — "thing you didn't know"

An 8-second vertical close-up of a single ice cube splitting cleanly along its
internal cracks under warm soft light. Slow zoom-in. Shallow depth of field,
cinematic realism. 9:16, 8s.

Audio: subtle crack sound, no music.
Negative: melting too fast, extra cubes appearing, frosted-glass look.

Tutorial intro — over-the-shoulder

An 8-second over-the-shoulder shot of a creator at a clean wooden desk, hands
typing on a mechanical keyboard, a monitor in soft focus showing colorful UI.
Warm desk lamp, slight handheld feel. 9:16, 8s.

Negative: extra fingers, doubled cursors, melting keyboard keys, illegible UI.

What goes in the negative prompt

The negative prompt does about half the work. A short, reusable default:

Avoid: floating objects, broken reflections, unrealistic gravity, jittery
motion, inconsistent shadows, distorted geometry, flickering textures,
unstable camera movement, melting faces, extra limbs, doubled letters,
warped text, identity drift between shots.

Use this as a baseline and add scene-specific items on top.

How Omni Flash builds these for you

Writing prompts by hand is fine if you only do one a week. For creators who ship every day, Omni Flash is designed to do the structuring work for you: paste a one-line idea, pick a workflow, and the studio outputs a final video prompt, scene breakdown, physics notes, character consistency notes, camera movement, visual style, lighting, audio direction, and a negative prompt — formatted for the workflow you picked.

FAQ

Do these prompts work with Veo 3.1 too?

Yes. The shot-list structure is model-agnostic. You can paste the same prompt into any text-to-video tool — the better the model, the more of the prompt it actually obeys.

Should I include an aspect ratio in the prompt?

Yes. Even when the platform exposes aspect ratio as a separate field, stating it in the prompt reinforces composition. Cinematic 16:9 prompts cropped to 9:16 often look bad.

Why does the character block matter?

Consistency models match on description, not on identity. The more your description stays the same across shots, the better your chance of "this looks like the same person".

How long should a prompt be?

Long enough to spell out the dimensions that matter; short enough that every line is doing work. Most great prompts are 80–200 words.

Is Omni Flash affiliated with Google?

No. Omni Flash is independent and is not affiliated with Google. Gemini, Gemini Omni, Omni Flash, Veo, and related names are trademarks of their respective owners.

Next steps