The best AI video prompts, 22 to copy (2026)

A good AI video prompt describes one visible action, the shot and camera movement, the light, the duration and what must not change. With that information the tool has less to guess, and you can iterate by changing one variable at a time. This library has 22 ready-to-copy video prompts for ChatGPT, Gemini and Copilot.

Add Founderz as a preferred source

What you’ll take from this

These are the four blocks you will find, built to take you from a loose idea to a clip you can review and edit.

• Prompts that ask ChatGPT, Gemini and Copilot to turn a brief into a video prompt, a shot list or a storyboard.
• Text-to-video and image-to-video templates for ads, B-roll, narrative scenes, social media and corporate video.
• The method behind each prompt: the fields a good instruction decides and the common mistakes to avoid.
• How to adapt the same prompt to ChatGPT, Gemini, Copilot, Claude and to generators like Runway, Kling or Veo.

If you want to work with AI video and images on real projects, the AI innovation program at Founderz is the natural next step. To build a custom prompt in a minute, use the free video prompt generator.

ChatGPT prompts for AI video

ChatGPT writes, organizes and fixes the prompt, the script and the shot list. To generate the clip you need an AI video generator, and what ChatGPT offers for that depends on your plan and version, so check your account. The text in brackets is an example: replace it with your case.

Turn an idea into a video prompt

Act as a video production director. I will give you a brief and I want you to turn it into a prompt for an AI video generator.
Brief: [a six-second vertical clip to introduce a vitamin C serum on social media].
Write one prompt with these parts, in this order: subject, visible action, shot and camera movement, light and style, duration, format and what must stay the same.
One action only, no on-screen text and no generated logos. Give it to me as one paragraph I can copy and paste, and add a shorter version below it.

Fix a video prompt that is not working

This video prompt is not giving me the result I want. Analyze it and rewrite it.
Current prompt: [paste your prompt here].
What went wrong: [the camera moves too much and the product changes shape].
Point out which instructions contradict each other or are unnecessary, and return a version with one action, one camera movement and a final sentence about what must not change.
Change only what is needed to fix the problem.

Prompt to create a video from an image

I am attaching an image that I will use as the first frame in an image-to-video tool. Describe it in three lines to confirm you understood it.
Then write a prompt that describes only the motion: what moves, how the camera moves and how long the clip lasts ([5 seconds]).
Do not repeat what is already visible in the image. End with one line saying what must stay intact: [face, product, brand colors and background].
Give me two variants: one with a locked camera and one with a slow lateral move.

Improve the quality of an AI-generated video

I generated a clip with AI and the quality is not what I need. Help me fix the prompt, not the clip.
Prompt used: [paste the prompt]. Tool: [Runway, Kling or Veo]. Flaws I see: [blurry edges, deformed hands, flicker between frames].
For each flaw, suggest one concrete prompt change (a single variable) and one tool setting I can check, such as duration, format or reference image.
Order the changes from most to least likely to fix the problem.

Shot list for a short video

Turn this brief into a shot list for a [30-second] video in [9:16] format.
Brief: [introduce our new booking app to small restaurants].
Return it as a table with these columns: shot number, role (opening, detail, demo, close), single action, shot type, camera movement and duration in seconds.
Maximum [6] shots. Each shot must work as a separate clip. Leave on-screen text and voice-over out of the shots, to add in editing.

Gemini prompts for AI video

Gemini accepts images and documents in the conversation, so it works for preparing prompts from your own material. If your plan includes video generation with Google models such as Veo, you can ask it for the clip; if not, copy the prompt into the tool you use. These three prompts take advantage of Gemini working with your photos and Google documents.

Video prompt with camera, light and sound

Write me a video prompt ready to paste into an AI video generator, with this structure: scene, action, camera, light, style, sound and duration.
Idea: [a café at sunrise where a barista serves an espresso].
Medium shot with a slow dolly in, warm side light from a window, clean documentary style, [eight seconds], horizontal 16:9 format.
Sound: [soft café ambience, no music or dialogue], in a separate sentence in case the tool supports audio.
No on-screen text or logos.

Animate a photo from the attached image

Use the attached photo as the first frame. Write the motion prompt to animate it: [the person blinks, smiles slightly and turns their head a few degrees toward the camera].
Locked camera, [5 seconds], calm pace. Keep facial features, clothing and background identical.
If you can generate the clip with my version, do it. If not, give me only the prompt to copy into Veo, Runway or Kling.

Script and shots from a Google Docs document

I am pasting the content of a document with the script for a [workplace safety training] video. Read it and return a table with one shot per scene: what is visible, what happens, shot type and duration.
Then write a video prompt for shot number [1], with one action and the elements that must stay the same.
Mark the parts of the script that cannot be shown with an image and should stay in the voice-over.

Copilot prompts for AI video

Copilot works inside Word, PowerPoint and Teams with your organization’s files, which fits the preparation of a video: brief, script and storyboard. The clip itself is generated afterward in a video generator. All three prompts start from a document, a presentation or a meeting you already have.

Storyboard table from a Word brief

Use the Word document [Spring campaign brief] as the source. Create a storyboard as a table with one row per shot: number, what is visible, action, camera, duration and continuity notes (product, wardrobe, brand colors).
Maximum [8] shots for a [45-second] video.
Use only the information in the document. If something is missing, list it as an open question at the end.

30-second script from a PowerPoint deck

Use the presentation [Quarterly results] to write the script for a [30-second] video for [the intranet].
One voice-over sentence per key slide, maximum [5]. For each sentence suggest which image or shot would go with it and how long it lasts.
Tone: [friendly and professional]. Do not add figures that are not in the presentation.

Short clips from a Teams meeting

Using the transcript of the Teams meeting [Product launch, October 12], suggest [3] short clips of [20 seconds] to announce the launch to [the whole team].
For each clip give the main message, the shots that would need to be generated or filmed and one voice-over sentence.
Flag which parts contain confidential data or customer names and should not be used.

More AI video prompts for generators like Runway, Kling or Veo

These prompt templates go straight into the video generator. Each one separates subject, action, camera, look, duration and constraints, and you can customize what is in brackets. For a long sequence, write one prompt per shot and decide the order and the sound in editing.

Text to video: ads, B-roll and narrative scenes

With text to video there is no visual reference yet, so the prompt has to set the first shot, the natural or studio light, the palette and the cinematic style you want.

Vertical product ad

Create a vertical 9:16 ad of six seconds. Macro close-up of [a glass serum bottle] on [wet light stone].
The bottle slowly turns a quarter turn while small water drops slide down its surface.
Slow dolly in, soft studio light, warm earth tones, clean background and shallow depth of field.
Keep the product identical throughout the clip. Leave negative space at the top for text added in editing.
No legible text, logos or generated watermarks.

B-roll for an explainer video

Horizontal 16:9 medium shot of [a professional reviewing a packaging prototype] on a bright work table.
Single action: [she turns the package and writes a short note in a notebook].
Static camera, soft natural morning light, clean documentary style, [five seconds].
Keep the notebook out of focus to avoid legible text. Leave clear space on the left for titles added in editing.

Narrative shot for a cinematic piece

Wide shot at sunrise of [a cyclist] arriving at [a coastal viewpoint], leaving the bike by a railing and looking at the sea.
Smooth lateral tracking shot, cool early light, light mist, calm pace, [eight seconds].
One main action and realistic physical movement. No cuts, no on-screen text and no wardrobe or identity changes between frames.

UGC-style clip for social media

UGC-style clip of [a person in their kitchen trying a protein shake]. Medium shot at eye level, handheld camera with slight movement, natural window light, phone-recorded look.
Single action: [picks up the glass, tastes it and nods].
Duration of [six seconds], 9:16 format. No on-screen text, no generated logos and no clothing changes between frames.

Image to video: animate photos and existing campaigns

With image to video, the reference already carries most of the art direction. So the prompt focuses on the motion and on what stays protected, without describing again what is visible.

Animate an existing campaign

Use the reference image as the visual source of truth. Keep exactly [product, composition, brand colors, wardrobe and background].
Add only a very slow lateral camera move and a slight believable physical movement in [the hair and the fabric].
Duration of [four seconds], calm pace, stable continuity between frames. Do not introduce new elements or change the character's expression.

Animate a portrait

Use the photo as the first frame. [The person blinks, smiles slightly and turns their head a few degrees toward the camera].
Locked camera, slight hair movement from a breeze. Keep facial features, clothing and background identical. Duration of [five seconds].

Animate a product photo

Use the photo as the first frame. Slow camera push toward [the product], with a light reflection that travels across the surface from left to right.
The product, the label and the colors do not change. Vertical 9:16 format, [four seconds], calm pace.

Animate a scenery or travel photo

Use the photo as the first frame. Clouds drift slowly, the water moves with gentle waves and the camera makes a very slow pan to the right.
Do not add people or new objects. [Six seconds], no cuts.

Social media, corporate video and presenters

For a social video, decide what must be understood in the first second and which area stays free for captions. In corporate video, turn each step of the process into its own shot. If you work with a real person, get their permission first.

Reel with a visible transformation

Vertical 9:16 reel of [four] seconds with a visible transformation: [a messy desk that becomes tidy in a single camera move].
Overhead close-up, camera slowly moving down, soft natural light, brisk pace. The action must be clear in the first second.
Leave the top third free for captions. No generated text or logos.

Resource shot for corporate video

Resource shot for a corporate video about [the onboarding process at a logistics company]. Horizontal 16:9 medium shot of [two people reviewing a tablet in a tidy warehouse].
Single action: [one points at the screen and the other nods]. Static camera, clean neutral light, documentary style, [six seconds].
No legible text or logos. Leave free space on the right for a title added in editing.

AI presenter or avatar

Script for an AI presenter: [Hi, I am Marta. In 30 seconds I will show you how to book a meeting room].
Framing: chest-up shot, presenter standing, background [a bright, simple office], wardrobe [a blue shirt]. Language: [US English].
Natural gestures, looking at the camera, no sudden movements. Check lip sync and the pronunciation of [the room name] before using the clip.

How do you write a prompt to create AI videos?

You write it as a production decision: what role the clip plays, what is visible, what happens, how the camera moves and what format it needs. These are the seven fields every prompt in the library decides, with an example of each.

Length is not the criterion. A simple shot works with one short sentence, and a sequence needs more detail but still a clear priority. If you ask for two incompatible styles, three actions and several camera changes in a few seconds, the model has no reason to resolve them the way you expect.

How the prompt changes in ChatGPT, Gemini, Copilot and Claude

Each tool helps with a different part of the process. These are the differences that matter when you copy a prompt from the library.

• ChatGPT: understands full sentences and accepts images, so it works for turning a brief into a prompt, fixing it and preparing the shot list. Ask for one variable per change.
• Gemini: accepts photos and documents and, depending on your plan, can generate clips with Google models. Put what must stay the same first and the motion second.
• Copilot: works with your Word, PowerPoint and Teams files. Use it for the brief, the script and the storyboard, and ask it to stick to the information in the document.
• Claude: does not generate video. It is useful for carefully drafting a long brief or a script and for writing the prompt, which you then paste into a video generator.
• Runway, Kling and Veo: each generator has its own controls for duration, format and reference. Adapt the prompt to what is available that day instead of copying it unchanged.

Common mistakes when creating AI videos

Most failed clips come down to one of these six production decisions that the prompt did not settle.

• Asking for a full piece in a single clip: split the work into shots when there are several actions or location changes.
• Confusing more words with more control: remove the details that do not affect the shot.
• Not separating text to video from image to video: with a reference image, describe mostly the motion.
• Forgetting the target format: the framing should be built for the final aspect ratio, not depend on a crop.
• Asking for final text inside the shot: add headlines, prices and legal messages in editing.
• Not checking rights and data: use only references, voices and material you have permission for, and do not upload confidential information without reading the tool’s terms.

A workflow for generating AI videos

The most reliable way to produce with AI is to treat each generation as one shot in a process, not as the finished piece. This workflow reduces repeats and makes clear what a person reviews.

1. Define the piece: audience, message, channel, format, duration and a success criterion you can check.
2. Split the idea into shots, each with one main action and one role: opening, detail, demo or close.
3. Prepare the references you have permission to use: product, palette, composition or a starting image.
4. Generate a first version with the base prompt and without changing five parameters at once.
5. Evaluate against visible criteria: fidelity to the brief, continuity, camera, hands, text, brands and safe area.
6. Iterate by changing one variable, save the prompt that worked as a template and add captions, music and voice-over in editing.

To sharpen the general method, the guide on how to ask AI questions works as well for a text model as for a video generator.

Prompt structures and camera work for better results

Prompt engineering for video starts with a simple text prompt: one subject, one action and one setting. Using simple text prompts first shows you what a model does with little guidance, and then you refine the visual style, the camera angles and the mood of your video one change at a time. Most of these prompts work in different AI tools, from Runway and Kling AI to Adobe Firefly and Veo, and these proven prompts give better results than a long list of adjectives. A good prompt is short, specific and easy to check against the clip, and an AI prompt written this way keeps your video content flexible so the creative possibilities stay open for the next iteration.

• Camera direction: say whether the camera is smooth, locked or handheld. Camera control changes the most from one generator to another, so a smooth camera push-in is safer than a complex orbit. Describe the camera motion and the camera work in one short sentence, and open with an establishing shot when the scene needs context.
• Animation and facial expressions: for animation, ask for one small movement. A slight smile or a blink reads well, but several facial expressions at once usually distort the face.
• Text overlays: do not ask for an overlay inside the clip. Add text overlays, prices and legal lines in editing, where you control the spelling.
• Duration limits: every tool has duration limits per generation, so plan short video clips and join them in your editor.
• Brand consistency: use the same reference image, palette and wording in every shot so the product looks the same across the piece.

These use cases cover a marketing video, a demo video, social media and corporate video. For a marketer, a text prompt is the quickest way to test a cinematic scene, and an AI image to video approach adds control because the image fixes the product. Browse the video examples above, copy the image to video prompts and adapt the image prompts to your own brand to present a product or to start creating high-quality video for your creative projects. Cinematic video and cinematic AI results depend more on the prompt than on the tool, which is why creative prompts with a clear mood and a single action tend to win. If you open Kling or any other generator, check the current limits first, because Kling video 3.0 and its rivals change their options often. These effective prompts and prompt structures are a good base for video creation in 2026: prompts for video generation in 2026 reward simple text and clear camera direction. Keep a short video workflow, save the best prompts for AI that worked, and treat each clip as a draft you refine.

Frequently asked questions about AI video prompts

These are the most common questions when you start writing video prompts, with the answer first and the library above to copy the examples from.

What prompt do I use to create AI videos?

One that describes a single visible action, the shot, the camera movement, the light, the duration and what must not change. For example: “Close-up of a bottle that turns a quarter turn, slow dolly in, soft studio light, 9:16, six seconds.” You have full templates in the library.

How do I write a prompt to create a video from an image?

Upload the image as the first frame and write only the motion: what moves, how the camera moves and how long it lasts. Do not repeat what is already visible. Add one line about what must stay intact, such as the face, the product or the brand colors.

How do I make videos in ChatGPT?

ChatGPT helps you write the prompt, the script and the shot list. To generate the clip you need an AI video generator, and what ChatGPT offers for that depends on your plan and version. If it does not include one, paste the prompt into Runway, Kling or Veo.

What prompt improves the quality of a video?

A prompt cannot improve a clip that already exists: fix the prompt you generated it with. Ask ChatGPT to suggest one change for each flaw, such as blurry edges or flicker, and check the duration, format and reference image. Change one variable at a time.

Can I use Claude to make AI videos?

Claude does not generate video. It does help you carefully draft a long brief, a script or the shot list and write the final prompt. Then you paste that prompt into a video generator such as Runway, Kling or Veo.

Can Gemini generate videos from a prompt?

Depending on your plan and country, Gemini can generate clips with Google video models such as Veo. Check what your version includes in your account. If not, use it to prepare the prompt from your photos and documents and paste it into another generator.

Is Copilot useful for writing video prompts?

Yes, for the preparation. Copilot works with your Word, PowerPoint and Teams files and can turn a brief or a presentation into a script and a storyboard. The clip is generated afterward in a video generator, with the prompt Copilot helped you write.

What should I include in a prompt for a video ad?

The role of the ad, product and action, shot, camera movement, light, duration, format and product continuity. Ask for free space for text and do not ask for legible text inside the clip: headlines, prices and calls to action are added in editing.

Your next step: generate, review and keep what works

Pick a piece you need this week, copy one of the prompts, change what is in brackets and generate a first version. Then change a single variable, such as the camera or the duration, and keep the prompt that worked as a template for your ads, product clips, B-roll and social content.

If you want a fuller audiovisual creation workflow, take a look at the AI innovation program at Founderz.