Blog / 9 min read · 2026-10-14

How to Write a Gemini Video Prompt: Formula, Examples, and Copy-Paste Templates

Learn how to write a Gemini video prompt with a practical formula, shot directions, audio cues, and copy-paste examples for better video results on demand.

To write a Gemini video prompt, describe one clear shot: who or what appears, what happens, where it happens, how the camera moves, and what viewers should hear. Add style and lighting only when they support the idea. Gemini Apps can generate videos from a text prompt and can also use uploaded images or video as inputs, while Google’s video guidance recommends clear descriptions of subject, action, style, camera, composition, focus, and ambiance. Google Gemini Apps Help Google AI for Developers

The Gemini video prompt formula

Use this reusable formula:

[Shot and subject] + [specific action] + [setting and time] + [camera movement] + [visual style, light, and composition] + [audio or dialogue] + [one constraint]

You do not need every part for every video. A simple product clip may only need a subject, action, camera move, and lighting. A narrative scene benefits from sound and a precise emotional beat.

Here is the formula in practice:

Close-up of a barista pouring latte art into a ceramic cup. Morning light fills a small, quiet café. The camera slowly pushes in from table height. Warm editorial food-film look, shallow depth of field, steam visible against the window. Soft espresso-machine hiss and low café ambience. Keep the cup centered; no text on screen.

The useful details are doing separate jobs:

Google’s Veo prompt guidance specifically supports cues for dialogue, sound effects, and ambient sound. Put exact spoken words in quotation marks; describe other sounds plainly. Google AI for Developers

Start with a single shot, not a vague idea

“Make a cool travel video” leaves nearly every decision open. Turn it into a moment the camera could actually capture.

Weak prompt:

Make a cool travel video in Japan.

Working prompt:

Wide establishing shot of a lone traveler crossing a narrow street in Kyoto just after rain. Paper lanterns glow along wooden storefronts, and reflections shimmer on the pavement. The camera glides forward at walking pace behind the traveler, then gently tilts up toward the lanterns. Calm, cinematic night photography with deep blue shadows and warm amber highlights. Light rainfall, footsteps on wet stone, distant conversation. No text or subtitles.

A video prompt is easier to control when it answers these questions:

1. What is the first thing on screen?
2. What changes before the shot ends?
3. Where is the camera?
4. What should the soundtrack communicate?
5. Which one visual outcome matters most?

If you cannot answer the first two, simplify the idea before adding more adjectives.

Write camera direction that is easy to visualize

Camera terms work best when they match the scene. A macro shot belongs on a small object. A wide establishing shot helps locate a person in a place. Avoid stacking camera moves that fight each other, such as “static handheld drone shot.”

Useful camera language includes:

• Close-up: face, hands, food, product detail, or a small emotional reaction.
• Wide shot: setting, scale, or a subject moving through a location.
• Over-the-shoulder: a person looking at a phone, map, artwork, or another person.
• Low angle: makes a subject feel imposing or dramatic.
• Top-down: useful for a desk, recipe, layout, or patterned movement.
• Slow push-in: builds attention toward a face or object.
• Tracking shot: follows motion alongside or behind a subject.
• Pan: reveals information across a scene.
• Rack focus: shifts attention between foreground and background.

Use one primary move. If the action needs a reveal, add one secondary move near the end.

Copy-paste template: cinematic character moment

Medium close-up of [character description] standing in [specific location]. [Character] slowly [single action], then pauses as [small reaction or reveal]. The camera [primary movement] at [camera height or angle]. [Style] with [lighting and color detail]. [Ambient sound and one sound effect]. Keep the focus on [important visual priority]; no on-screen text.

Example:

Medium close-up of an elderly watchmaker in a narrow workshop filled with brass gears. He carefully closes a pocket watch, then smiles when it begins ticking. The camera makes a slow push-in from eye level. Quiet period-drama realism, warm window light, floating dust, rich brown wood and brass tones. Clock ticking, faint street rain, a soft wooden drawer closing. Keep the watchmaker’s hands and the watch in sharp focus; no on-screen text.

Add action that can happen within one short clip

A prompt often fails because it asks for a whole story: arrival, conflict, chase, resolution, and a final speech. Build one beat instead.

Choose actions with a visible beginning and end:

• A cyclist rounds a corner and enters sunlight.
• A chef lifts the lid from a steaming pot.
• A child releases a paper boat into a stream.
• A robot pauses, turns toward a blinking signal, and reaches for it.

This matters especially for clips built from an image. If your starting image already establishes the character and location, write the motion and camera direction rather than repeating every visual detail. Google’s Veo documentation says an input image can serve as the initial frame for image-to-video generation. Google AI for Developers

For more scene ideas built around a supplied picture, see these Gemini image-to-video prompt examples.

Copy-paste template: product video

Clean studio product video of [product] on [surface]. The product [single physical action or interaction]. Start with a [shot type], then [camera movement] to reveal [key feature]. [Lighting direction], [background color], and [visual style]. [Relevant sound effect or music mood]. Keep the product shape, color, and label area clear; no extra objects or readable text.

Example:

Clean studio product video of a matte black insulated water bottle on pale stone. A bead of condensation rolls down the side as the bottle slowly rotates. Start with a close-up of the cap, then make a slow orbit to reveal the full silhouette. Soft side lighting, charcoal-gray background, premium minimal commercial style. Gentle water droplet sound and restrained electronic pulse. Keep the bottle shape and matte texture clear; no extra objects or readable text.

Prompt audio deliberately

Audio should reinforce the image, not compete with it. Describe the layer you need:

• Ambiance: “waves and distant gulls”
• Sound effect: “a zipper closes with a crisp fabric sound”
• Music direction: “slow, sparse piano”
• Dialogue: “She says, ‘We made it,’ in a quiet, breathless voice”

For dialogue, name the speaker and emotional delivery so the line has context. Keep spoken lines short enough to fit the action.

Copy-paste template: dialogue scene

[Shot type] of [speaker] in [location]. [Speaker] [action] and says, “[short line],” in a [tone] voice. [Second character, if needed] responds with [reaction]. The camera [movement]. [Lighting and style]. Audio includes [ambient sound] and [sound effects]. Prioritize natural pauses and clear facial expressions.

Example:

Two-shot at a kitchen table during a midnight power outage. A woman lights a candle, looks at her brother, and says, “We can start over tomorrow,” in a tired but steady voice. He exhales and nods. The camera holds still, then slowly pushes closer as the flame steadies. Intimate indie-film realism, soft candlelight, deep surrounding shadows. Rain taps the window; the match flares, then fades. Prioritize natural pauses and clear facial expressions.

Use image, video, and editing prompts differently

In Gemini Apps, you can add images or a video to a video-generation prompt, then ask for edits such as changing camera angles, modifying a scene, or removing or replacing objects. Google notes that available upload and editing options can vary by region. Google Gemini Apps Help

When providing a reference image, do not write a generic text-to-video prompt. State what must stay consistent and what should move.

Use the uploaded image as the opening frame. Preserve the woman’s red coat, hairstyle, and the foggy train platform. She turns toward the arriving train as wind lifts the edge of her coat. The camera tracks from her left side, ending on a close-up of her expression. Cool morning light, realistic film grain, distant rail screech and low station announcement. Do not add other people in the foreground.

For an edit, name the change and preserve the rest:

Keep the subject, framing, and action unchanged. Replace the cloudy afternoon sky with a clear golden-hour sky. Add warm sunlight from camera left and subtle long shadows. Preserve the original camera motion and ambient city sound.

Common Gemini video prompt mistakes

Listing styles without a visual subject

“Cinematic, viral, 4K, aesthetic, dramatic” does not tell the model what to show. Begin with the object, person, or event. Style comes after the scene is understandable.

Giving contradictory instructions

“Dark noir scene with bright midday sunlight” can work only if you explain the contrast, such as harsh sun entering a dark interior. Check lighting, time of day, color palette, and mood for conflicts.

Trying to control every second

Describe the most important starting image, action, and ending beat. Let the camera motion connect them. Overly choreographed prompts are harder to revise because you cannot tell which instruction caused the unwanted result.

Forgetting the soundscape

If the scene needs audio, say so. A bakery without the scrape of a tray, a forest without wind, or a dialogue scene without an emotional tone can feel unfinished.

FAQ

How long should a Gemini video prompt be?

Long enough to identify the subject, action, setting, camera, look, and sound. For a straightforward clip, one focused paragraph is usually clearer than a long screenplay.

Should I use full sentences or keywords?

Use natural, descriptive sentences. Keywords are useful inside those sentences for camera terms, lighting, and style, but the action should read as a clear sequence.

Can I ask Gemini to create dialogue and sound effects?

Yes. Google’s video prompt guidance documents dialogue, sound-effect, and ambient-noise cues. Put exact dialogue in quotation marks and describe the delivery and surrounding sounds. Google AI for Developers

How do I revise a video that is close but not right?

Change one variable at a time: camera movement, lighting, action, or audio. For example, keep the scene prompt and replace “handheld tracking shot” with “locked-off wide shot.” Gemini Apps also supports prompt-based edits to generated videos. Google Gemini Apps Help

Final prompt checklist

Before submitting, check that your prompt has one visible subject, one main action, a camera viewpoint, a setting, and only the audio cues that matter. If a result misses the mark, revise the single instruction tied to that problem instead of rewriting everything at once.

Related guide

For a photo-based scene collection, see the couple video prompt guide.