AI Video Generation Prompts for Google Flow, Grok Video and Kling AI

Good AI video generation prompts describe a subject, one clear action, a setting and a camera view. Add lighting and sound where the selected model supports them. For Google Flow, start with the 15 original Veo 3 scene prompts below. For Grok Video or Kling AI, use the separate examples and adapt them to text-to-video, image-to-video or reference-led generation.

This guide contains 31 prompt ideas: 15 original Google Veo3 prompts for Flow, eight Grok video prompts and eight Kling AI prompts. It also explains how to transfer the same idea between tools without confusing a text description with a setting in the interface. The new examples are writing templates, not claims of tested outputs or guaranteed results. The original Veo prompts and their existing video embeds are retained unchanged.

Which tool should you use for which prompt?

Choose by your workflow, not by a claim that one model is always best. These are practical starting points, not a head-to-head performance benchmark. Model, plan and regional availability can change; use the controls actually offered in your account.

Tool	Start here when you want	What to put in the prompt
Google Flow with Veo	A cinematic scene with a planned camera view, atmosphere and audio cues	Subject, action, location, framing, light, style and concise sound or dialogue instructions. Use references or frames only in modes that support them.
Grok Video / Grok Imagine	A text-led scene, an animated still, or a reference-led video workflow	Describe visible action and what should change. With an input image, focus on motion rather than rewriting its appearance. Set technical controls separately where available.
Kling AI	A subject-and-motion scene or image animation with explicit movement	For text-to-video: subject + movement + scene, then camera, lighting and atmosphere. For image-to-video: subject movement + background movement.

How to write an AI video prompt that is easy to follow

A prompt is a short shot brief. It should tell the model what the audience sees, what happens and how the scene is filmed. An adjective such as “cinematic” cannot replace the subject or the action. Begin with a concrete sentence: “A ceramic cup rests on a wooden counter while steam rises.” Then add the frame: “Close-up, locked camera, warm window light.”

Reusable template: [subject and identifying details] + [one action] + [setting] + [camera view or movement] + [lighting and visual style] + [sound or dialogue, if supported]. This is an editorial template assembled from the vendors’ guidance, not a special command syntax.

• Keep the main action small enough to fit the chosen clip. A person opening a door is a clearer starting task than entering a city, meeting three characters and escaping in the same shot.
• Choose one camera instruction initially. “Locked wide shot” and “rapid handheld orbit” describe different shots; do not ask for both at once.
• Separate the picture from the soundtrack. Say what makes a sound, what the background ambience is and whether someone speaks. A model can still miss an instruction.
• Set length, aspect ratio, resolution and other technical options in the tool when controls exist. Writing “4K” in a paragraph does not create a 4K export option.
• For image-to-video, inspect the input image first. Ask visible elements to move rather than asking the model to replace the whole scene.
• Keep a record of the prompt, model/mode, input asset and settings. Change one variable per revision so you can tell what helped.

Free Veo 3 prompts, Veo 2 examples and community ideas

The prompt text on this page is free to copy and adapt. That does not mean generation in Flow, Grok or Kling is free: access, usage limits and credit charges depend on your account and selected model. Check them before generating.

For Veo 2 prompt examples, reuse the visual subject, action, scene and camera brief, but do not assume the audio or dialogue instructions in a Veo 3 prompt apply to Veo 2. Check the mode’s supported features. A good visual brief can transfer between versions; a feature-specific instruction may not.

Searching for Veo3 prompts on Reddit can help you find creative ideas, but a posted clip is not evidence that a prompt will reproduce the same result in your account. Check the creator’s model, settings, reference images and edits. Do not treat a username or an unexplained community term as a product feature.

The examples below are in English. If you want Polish, Italian or Russian Veo prompts, translate the shot brief while retaining its subject, camera and action details. Evaluate any dialogue separately; translating the prompt is not a guarantee of supported speech or subtitles.

Google Flow prompts: use the original 15 Veo 3 scene ideas

Google Flow is Google’s filmmaking workspace; Veo is a video generation model used within it. “Google Flow prompts”, “Veo 3 prompts” and “Google Veo3 prompts” often refer to the same creative task, but the workspace and the model are not interchangeable names. Check the selected model and supported mode before you generate.

Google’s Veo prompt guide emphasizes framing and motion, visual style, lighting, characters, location, action and dialogue. Its Flow help explains that a text prompt should complement rather than contradict supplied frames or ingredients. These principles are more useful than relying on old screenshots or assuming every model has the same controls.

How to use the preserved examples: open a Flow project, choose an available Veo mode, paste a prompt and review the output. The more elaborate examples contain several actions or audio instructions. If the result loses the main idea, try a shorter duplicate prompt that isolates one shot. The original text below is intentionally unchanged; it is a creative brief, not a promise that all actions or exact dialogue will appear.

The embedded clips belong to the original article. They are not presented as new tests of the current Flow interface. We have not rerun these prompts across the three tools.

---

1. Rain-Slick Drift Chase

A black coupé hydroplanes through neon puddles while police lights bounce off skyscraper glass.

Veo3 Prompt: Drifts through neon-lit, rain-washed streets, a black coupé fishtails past storefronts while police cruisers blaze in hot pursuit. A low dolly hugs the spinning wheels; sparks arc as the car skims a corner. Inside, a drenched driver fixes a steely gaze on the road. Sirens slice the air, fused to an urgent electronic score.

2. Normandy Ghost Walk

Shell craters, cold rain, a soldier’s hollow stare—World War II in slow motion.

Veo3 Prompt: Tracks a lone American soldier across a cratered battlefield at dusk, handheld and facing him in reverse. Cold rain pellets his hollow, mud-streaked face; artillery flashes silhouette broken trees behind. He halts, sinks to his knees, and whispers, “Why am I still here?” A mournful orchestra swells beneath the thunder.

3. Android Awakening

Steel table, flickering fluorescents, one sudden blue-eyed blink that changes everything.

Veo3 Prompt: Dark lab interior. Close-up on a lifeless android’s face. Its eyes suddenly glow blue and snap open. Camera pulls back slowly as steam rises from the table. Flickering lights. A deep electronic hum builds. The android turns its head toward the camera in silence

4. Viking Last Stand

Snow drifts through blazing thatch while a fur-clad warrior screams at the sky.

Veo3 Prompt: A Viking warrior stands in the snowy ruins of a burning Nordic village, wearing fur and leather. He grips an axe in his hand. Snow falls through rising smoke. He roars to the sky. The camera slowly circles him, then zooms in on his face. A deep Nordic choir and tense strings swell

5. Alley Panic SnorriCam

Camera locked to the runner’s face—every gasp, every jolt, pure adrenaline.

Veo3 Prompt: Night. A terrified figure runs through a narrow, dark alley. The SnorriCam is mounted to their torso, locked on their face—eyes wide with fear. Every step jolts the frame. Heavy breathing and echoing footsteps follow. A tense electronic score builds with each frantic second

6. Cliff-Edge Destiny

Cloak snapping in the storm above a raging sea—the hero before the leap.

Veo3 Prompt: A lone person stands on a cliff above a wild sea. Wind pulls at a dark cloak and long hair. A sword rests on their back. The camera walks up behind them, then circles to show the endless water ahead. Waves smash against the rocks below. Deep strings and a rising choir fill the air, building to a powerful moment.

7. Ash-Blown Homecoming

A survivor whispers to a dead skyline, memories drifting like dust.

Veo3 Prompt: A lone survivor stands on a rusty bridge, looking at a shattered city. His flannel shirt is torn, jacket scuffed, boots thick with mud. His beard is rough and his eyes look worn out. Grey ash drifts through the air. He whispers, “This was home.” The camera eases back, widening the view of the empty skyline behind him.

8. Inferno Bridge Motorcycle Jump

Flames roar, the bridge collapses, and the bike flies through fire in slomo.

Veo3 Prompt: Night. A woman in a black latex suit races her motorcycle over a collapsing bridge. Debris smashes down behind her, flames shoot into the sky. A low-angle camera glides beside the spinning wheels. She hits a shattered gap and sails through it in slow motion, framed by fire. Heavy drums pound louder with every second

9. Interrogation Lamp Confession

A flickering bulb, one steady stare, and a single chilling line.

Veo3 Prompt: A dim interrogation room. A single lamp flickers overhead, throwing shaky light across a metal table. A police officer stands with folded arms. Across from him, a suspect sits in handcuffs, still and silent. The camera holds in a steady mid-shot, then begins a slow push toward the suspect’s face. The air feels thick, broken only by a low, steady hum. At last he speaks, eyes fixed on the officer: “You already know what I did. The real question is why.”

10. One-Take SWAT Breach

Boots thunder up stairwells, flashlights carve the dark, tension never blinks.

Veo3 Prompt: Night. A SWAT team in black armor bursts through an apartment doorway, rifles up, lasers cutting through dust. Flashlights sweep the narrow hall, catching peeling paint and drifting smoke. Heavy boots pound the floorboards; every step echoes. The camera hugs the lead officer’s shoulder, slipping with the stack as they fan out through stairwells and tight corridors. Doors slam open one by one. Shouts—“Clear!” “Move!”—bounce off the walls. Breath comes sharp through mask filters. A fierce drumline in the score drives faster, matching the pulse of boot-beats and rattling hinges.

11. Warehouse Reborn

A static street-level camera watches a rusted warehouse refold itself—panels slide, steel turns to smart glass, and vertical gardens bloom—until it stands as a glowing, plant-wrapped living space at golden hour.

Veo3 Prompt: Cinematic wide shot from a static street-level camera, capturing a rusted, decaying warehouse. Gradually, the corrugated panels begin to retract and refold—steel skeletons pivot and slide, rust transforms into brushed concrete and smart glass, and lush greenery unfurls from vertical planters as the structure blossoms into an award-winning architectural living space with floating terraces, cantilevered staircases, and sculptural lighting. Style: Brutalist-to-biophilic kinetic metamorphosis, seamless architectural reassembly with sound design of creaks turning into ambient echoes, dusk-to-golden-hour lighting shift, texture-rich closeups enhancing metal-to-living-material transitions.

12. Submerged Spacewreck Discovery

Floodlights pierce emerald water, revealing a shattered starship hull on the ocean floor.

Veo3 Prompt: Drifts through murky depths toward a broken spacecraft; muffled sonar pings echo while a vocal pad hums in eerie harmony.

13. Monsoon Rooftop Dance

Rain sheets down neon signs as two strangers spin beneath thunder.

Veo3 Prompt: 360-degree orbit around dancers on a Bangkok rooftop during monsoon downpour; joyful house beat syncs with every splash.

14. Post-Apocalyptic Train Escape

Rusty freight cars barrel past desolate plains while rebels cling to the roof.

Veo3 Prompt: Low drone chase beside a speeding train through wasteland; distorted rock riff and wind roar intertwine.

15. Lantern-Lit Temple Showdown

Ancient stone glows under crimson lanterns as swords clash in silence.

Veo3 Prompt: Glides steady-cam between fighters inside misty temple hall; suspended taiko drum pulse punctuates each blade spark.

Using Veo prompts in Google Flow without inventing settings

For a rain-chase scene, the original prompt combines vehicle motion, a wheel-level camera, a driver close-up, sirens and music. If your clip cannot fit all that, keep the chase and low tracking shot for one generation, then make the driver reaction a separate shot. Save the original prompt as the starting idea; make edits in a working copy rather than losing it.

For dialogue scenes such as the interrogation example, keep speech brief and identify the speaker. Google’s guidance explicitly discusses voice, action and dialogue, but it does not promise perfect speech or synchronization in every generation. Listen to the result rather than assuming the quoted line was spoken correctly.

For consistent characters, use available reference or ingredient features and follow the selected mode’s rules. Repeat essential appearance and wardrobe details in each shot brief. Reference inputs and prompt wording work together; neither is a guarantee of identical faces across every clip.

Use the actual controls shown in Flow rather than assuming a universal music, lip-sync or color-grading panel. A prompt guide is not an interface manual. Do finishing edits in your video editor when necessary.

Grok video prompts: describe the scene, then the change

xAI’s official video generation documentation distinguishes text-to-video, image-to-video, reference-to-video, editing and extension workflows. Its examples use concrete scenes and motion, including a cat in a sunbeam with a gently swishing tail. That is a sound basis for a prompt library: name the subject, define its motion and avoid unrelated scene changes.

Important interface distinction: the xAI documentation is for the API. Model names, configurable length, audio options and reference support there are not proof that the same controls appear in every Grok consumer app. These examples can be used as natural-language shot briefs; check the mode and options in your own Grok interface. This guide does not claim Grok is inherently “more conversational” than the other video tools.

Grok prompt template: [visible subject] does [specific motion] in [setting]. [Camera and lighting]. [What stays consistent]. Add brief sound instructions only if your selected model/mode supports them. For an uploaded image, replace the long appearance description with the visible subject and desired animation.

1. Coffee steam product loop

A white ceramic coffee cup sits on a walnut counter beside a window. Thin steam rises and curls slowly above the cup. Locked close-up camera. Warm morning light. The cup, handle and counter remain unchanged throughout the shot.

Use this as a simple text-to-video scene. With a real product photograph, describe only the steam and lighting movement; do not ask for a new logo or a different cup.

2. Rain on a neon street

A black coupe drives steadily along one rain-soaked city street at night. The camera tracks beside the car at wheel height. Neon reflections slide across puddles. Keep the same car in view for the entire shot. Rain and tire hiss, if audio generation is enabled.

This adapts the original chase idea into one action. Add a driver close-up in a separate clip rather than asking the camera to be outside and inside the car simultaneously.

3. Animate a portrait photograph

Animate the person in the uploaded portrait with one gentle blink and a slight turn toward the window. Keep the facial features, hairstyle and clothing consistent with the source image. The camera stays fixed. Background remains still.

Image-to-video starting point. Use an image you own or have permission to animate. Identity consistency is an instruction to review, not a guaranteed result.

4. Paper rocket launch

A handmade paper rocket stands on a cardboard moon set. It lifts slowly upward while soft cotton clouds drift below. Wide static shot, stop-motion visual style, warm studio light. Keep the handcrafted paper texture visible.

Name the material and movement explicitly. Start with the lift-off before adding a full space journey or multiple destinations.

5. Architecture reveal

A small timber cabin sits in a forest clearing at sunrise. The camera moves slowly forward along the path toward the open doorway. Leaves sway lightly. Keep the cabin design stable, with soft mist behind the trees.

Useful as a scene brief or an animation of an authorized architectural image. Avoid asking an exact design to rebuild itself unless a transformation is truly the goal.

6. Cat in window light

A ginger cat lies in a sunbeam on a blue cushion. Its tail swishes slowly, and one ear turns toward an off-screen sound. Locked medium shot, soft afternoon light, quiet room ambience if supported.

Inspired by the kind of small movement used in xAI’s documentation, not copied as a tested result. Keep the animal action simple before requesting complex walking.

7. Short presenter welcome

An adult presenter in a plain studio faces the camera and gives a small welcoming nod. Medium shot, even soft lighting, neutral background. If speech is supported, the presenter says: "Here is the idea." No scene change.

Keep spoken text short and check the output manually. Add accurate captions in your editor if exact wording is important.

8. Animate a landscape still

In the uploaded mountain-lake image, gentle ripples move across the water and a thin cloud drifts slowly behind the peaks. The mountains and shoreline remain fixed. No camera movement.

The source image already supplies the composition. Describe only water, cloud and camera behavior rather than inventing a new landscape.

Grok revision notes: what to change when a prompt misses

If an image animation redesigns the subject, remove extra style descriptions and concentrate on the desired motion. If a text scene introduces too much action, reduce it to a single subject and one verb. If you are using xAI’s API, configure supported output parameters separately instead of burying them inside the scene description. In a consumer app, use only controls the interface actually exposes.

Editing, extension and generation are different tasks. For editing an existing clip, describe the change you want rather than writing a whole new story. For extension, preserve the scene and give a next action that follows its final frame. Availability depends on the selected product and model; the API documentation should not be treated as a complete consumer-app walkthrough.

Kling AI prompts: subject, motion and scene first

Kling’s text-to-video guide gives this formula: subject (subject description) + subject movement + scene (scene description) + camera language, lighting and atmosphere. Subject, movement and scene are the essential parts. Its image-to-video guide instead focuses on subject + movement, background + movement, because the image already provides the scene.

These are useful foundations, not a reason to copy every historical setting in a guide. Kling’s quickstart examples refer to particular modes and clip lengths. Your currently selected model may offer different options. Check the model selector and available controls before following any duration or camera-control instruction.

Kling text-to-video template: [subject and visible details]. [One movement]. [Scene]. [Framing or one camera move]. [Lighting and atmosphere]. Kling image-to-video template: [visible subject] performs [movement]; [background element] moves [how]; [camera behavior].

1. Skincare product reveal

A frosted glass skincare bottle stands on a pale stone plinth. A narrow highlight moves slowly across the glass as the camera makes a gentle push-in. Close-up framing, soft studio lighting, calm clean background.

For a branded packshot, use your authorized product reference and review whether the label remains accurate. Add final text in an editor if readability matters.

2. Fashion walk

An adult model wearing a beige coat walks three steady steps along an empty stone walkway. Medium full-body framing. The camera tracks backward smoothly. Overcast daylight, minimal background, consistent outfit.

Keep the movement and scene explicit. If hands or fabric distort, shorten the motion and try a simpler camera view before adding more detail.

3. Animate an android image

The android in the supplied image opens its eyes once and turns its head slightly toward the camera. A faint wisp of steam rises behind it. Camera fixed. Keep the face, body proportions and laboratory layout unchanged.

An image-to-video adaptation of Android Awakening. The uploaded image defines the design; the prompt defines a small movement rather than a new robot.

4. Cliff-side atmosphere

A traveler in a dark cloak stands on a rocky cliff above the sea. The cloak flutters gently and waves move below. Wide shot from behind the traveler. Camera slowly pushes forward. Cold dawn light and mist.

A single-scene adaptation of Cliff-Edge Destiny. An orbit can be a separate variation, but combining a push, orbit and zoom at once makes diagnosis harder.

5. Food steam close-up

A bowl of vegetable soup rests on a wooden table. Thin steam curls upward while a spoon remains still beside the bowl. Close-up, shallow depth of field, locked camera, warm kitchen light.

For an existing food photo, ask only for the steam movement and camera behavior. Do not request a new bowl or a changed ingredient arrangement unless intended.

6. Forest aerial

Aerial wide shot above a pine forest at sunrise. The camera advances slowly over the treetops. A thin band of mist drifts through the valley. Warm light touches the upper branches. Smooth continuous movement.

The action is the camera travel and drifting mist, not a chain of different locations. Keep direction consistent throughout the shot.

7. Dancers in the rain

Two adult dancers make one slow coordinated turn on a rain-soaked rooftop. Wide framing, camera fixed, neon reflections in puddles, rain falling behind them. Keep the pair centered and their clothing consistent.

An adaptation of Monsoon Rooftop Dance that simplifies the movement. Add a camera orbit only after the subject motion is acceptable.

8. Start-frame animation for a room

In the supplied living-room image, the curtain sways gently and sunlight shifts slightly across the floor. Furniture remains stationary. The camera stays locked in the original position.

Image-to-video fits when preserving a room’s existing composition matters. Treat any start/end-frame feature as a separate mode and use it only if offered by your selected model.

Camera movement in Kling: prompt language is not a control panel

A “close-up” specifies framing; a “slow push-in” specifies motion. Kling’s camera control guide makes that distinction and describes camera movement controls. Support can vary by mode. If the selected interface offers a camera control, make its direction agree with the text prompt instead of setting one direction and describing another.

Reference-led motion workflows also differ from ordinary image animation. If you supply a motion reference, do not contradict its action with an unrelated dance or walk in the text. Explain the scene or appearance change you want and follow that mode’s instructions. This article is not a claim that every Kling model offers identical motion controls.

The same scene adapted for Flow, Grok and Kling

The original Rain-Slick Drift Chase is useful for showing the difference between a scene idea and a mode-specific prompt. Its creative core is a black car, wet neon streets and a low camera. You can retain that core without promising identical outputs between tools.

Workflow	Adaptation approach
Google Flow / Veo	Use the original preserved prompt as the creative brief. If it is too busy, isolate the wheel-level chase and make the driver reaction a second shot. Keep audio instructions concise.
Grok text-to-video	Start with one car moving along one street, wheel-height tracking and rain ambience where supported. Configure any technical controls outside the descriptive prompt.
Kling text-to-video	Name the car, its movement and the street first, then framing, light and mood. Try one camera move rather than multiple competing directions.
Any supported image-to-video mode	Use an authorized image of the car and street. Describe the car motion, water movement and camera behavior, not a replacement vehicle or new location.

AI video prompt troubleshooting: symptom, likely issue, next test

These are prompt-editing suggestions, not measured success rates. A failed clip may also reflect model limits, safety restrictions, reference quality or unavailable features. Change one thing and review the result before spending more credits.

Problem	First thing to check	A practical next test
The subject changes appearance	Are text instructions contradicting the reference image?	Remove conflicting appearance details. Use a clearer authorized reference and restate only the desired motion.
The scene jumps or cuts unexpectedly	Does the prompt ask for several places or camera moves?	Ask for one continuous shot in one location and one camera direction.
Too many actions get skipped	Does the sequence fit the chosen clip length?	Split the action into separate clips. Start with one event and a reaction.
Speech is wrong or missing	Is audio supported, and is the line short enough?	Identify one speaker and one short line. Check the audio; finish exact narration or captions in an editor if needed.
Text or logos are distorted	Is the model being asked to generate precise lettering?	Use an authorized reference, review frame by frame and add exact graphic overlays after generation.
An image animates too much	Are style and scene changes overpowering motion instructions?	Name only a small visible motion with a locked camera; leave the input composition alone.
Export looks worse than expected	Did you confuse requested style with actual output settings?	Check the export options actually available. Do not assume a prompt can unlock a resolution.

A repeatable workflow for social clips, product videos and story scenes

• Write the purpose of the shot in one sentence. A product reveal needs readable product identity; a mood clip may only need texture and movement.
• Pick the generation mode before writing. A blank text-to-video scene needs the setting described; a supplied still already has one.
• Start with the simplest useful shot. Save its prompt and technical settings. Do not change camera, subject, lighting and action all together.
• Review motion, faces, hands, brand details, speech and scene continuity. A plausible thumbnail is not a complete quality check.
• Assemble longer stories from reviewed clips. Add final captions, exact titles, music and transitions in your editor where appropriate.
• Check rights and platform rules for input images, voices, real people and output use. Do not infer universal commercial-use permission from a prompt guide.

For more app and tool guides, browse Just Create App’s apps collection. Start with the tool that supports your inputs and workflow; a larger list of generators will not fix an unclear shot brief.

Frequently asked questions about AI video generation prompts