# AI video prompt examples: 30 prompts by format, matched to a model

> 30 AI video prompt examples in six formats, each matched to Veo 3.1, Kling 3.0 or Seedance 2.0 using Google, Kling and fal guides, with a first fix to try.

Canonical URL: https://honeyvid.ai/learn/ai-video-prompt-examples/
Updated: 2026-11-13
Publisher: HoneyVid (https://honeyvid.ai)
Topic: Prompts and publishing

## Short answer

A good AI video prompt names one subject, one action, one camera move and the sound, in the order your model's guide documents. Below are 30 prompts in six formats, matched to Veo 3.1, Kling 3.0 or Seedance 2.0 from guides read on 2026-10-26. Each has a first fix. If a clip drifts, change one element and rerun.

## Key facts

- **Veo 3.1 prompt formula (Google, 16 Oct 2025):** [Cinematography] + [Subject] + [Action] + [Context] + [Style & Ambiance] ([source](https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1))
- **Veo 3.1 audio syntax:** Dialogue in quotation marks; lines starting "SFX:" and "Ambient noise:"; timed shots written as [00:00-00:02] ([source](https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1))
- **Kling VIDEO 3.0 dialogue rule (Kling, 3 Jul 2026):** Name the speaker, write the line, describe the delivery; clips run 3 to 15 seconds ([source](https://kling.ai/blog/kling-video-3-omni-native-lip-sync-audio-guide))
- **Seedance 2.0 prompt shape (fal, 18 Jun 2026):** Subject and motion first; spoken lines in double quotes; write "no music" or the clip may come back scored; clips run 4 to 15 seconds ([source](https://fal.ai/learn/tools/seedance-2-0-prompting-guide))
- **YouTube disclosure rule:** Realistic AI content, including a real person shown saying or doing something they did not, must be disclosed ([source](https://support.google.com/youtube/answer/14328491))

Most pages with this title give a formula and a list of prompts. This page gives 30 prompts in six formats, names the model whose conventions each one follows, and puts a first fix beside every prompt. Prompts 26 and 27 adapt examples from Google's Veo 3.1 guide; the rest are original. The model matches come from the Google and Kling guides and fal's Seedance guide (fal hosts ByteDance's Seedance), read on 2026-10-26.

## What does a good AI video prompt contain?

A good AI video prompt contains one subject, one action, one camera move, the light and the sound. [Google's Veo 3.1 guide](https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1) orders these as cinematography, subject, action, context, then style and ambiance.

Seedance reverses the start. [fal's Seedance 2.0 guide](https://fal.ai/learn/tools/seedance-2-0-prompting-guide) says subject and motion do the heavy lifting and that mood words such as "beautiful" or "cinematic" give the model little to act on. [Kling's guide](https://kling.ai/blog/kling-video-3-omni-native-lip-sync-audio-guide) keeps speaker, line and delivery together and pairs acoustic details with the visual action. For ready-made slot templates, see the [AI video prompt generator](https://honeyvid.ai/learn/ai-video-prompt-generator/).

## Which model should each prompt format use?

Each format below leans on a feature a guide documents. The table shows which one, as of 2026-10-26.

| Format | Model used in this page | Documented feature behind the choice |
| --- | --- | --- |
| Product and ads | Veo 3.1 | Cinematography first; labelled SFX and ambient lines |
| Talking and dialogue | Kling 3.0 | Named speaker, line and delivery; 3 to 15 s |
| Short-form trends | Seedance 2.0 | Quoted lines, explicit "no music", 4 to 15 s |
| Camera moves | Veo 3.1 | Cinematography is the first slot in the formula |
| Image to video | Veo 3.1, Kling 3.0 | Start and end frame (Veo); reference elements (Kling) |
| Multi-shot stories | Veo 3.1, Seedance 2.0 | Timed shots (Veo); "cut to" in the prompt (Seedance) |

## Product and ad video prompt examples

The first three prompts use Veo 3.1's cinematography-first order. The last two use Seedance 2.0's subject-and-motion order. Replace the product with something you own, and read [the AI commercial workflow](https://honeyvid.ai/learn/ai-commercial/) for the full job.

1. **Bottle on a counter.** Suits Veo 3.1.
   > Slow push-in, medium close-up. A matte green glass bottle with a plain white label stands on a wooden kitchen counter. Morning window light, soft haze. Sound effects: a cap being turned. Ambient noise: a quiet kitchen.
   First fix: if the label text warps, remove the word "label" and keep the label blank in your reference.
2. **Perfume reveal.** Suits Veo 3.1.
   > Macro shot, slow orbit. A glass perfume bottle on black stone, a single spotlight, drifting mist. Dark, glossy, shallow depth of field. SFX: soft glass tap.
   First fix: if the bottle changes shape mid-orbit, shorten the move to a half circle.
3. **Sneaker on a turntable.** Suits Veo 3.1.
   > Low front angle, locked off. A white running shoe rotates on a slow turntable against a pale grey studio wall. Even soft light. Ambient noise: faint studio hum.
   First fix: if the shoe wobbles, say "turntable rotates steadily at one speed".
4. **Coffee pour.** Suits Seedance 2.0.
   > A hand pours black coffee into a white ceramic cup, steam rising. Kitchen at dawn, warm light. Tight single, slow push-in. Pouring liquid, a spoon clink, no music.
   First fix: if the clip gets music, repeat "no music" at the end.
5. **Skin cream, three cuts.** Suits Seedance 2.0.
   > A woman's hand opens a white jar on a bathroom shelf. Cut to a close-up of cream on a fingertip. Cut to the hand smoothing it on a wrist. Soft daylight. Quiet room tone, no music.
   First fix: if the cuts blur together, number the shots and cut the prompt to two.

Note that the exact claims of a real product, such as ingredients or results, belong in your edit, not your prompt.

## Talking and dialogue prompt examples

Kling 3.0 ties each line to a named speaker and a delivery note, so these prompts do the same. [Kling's guide](https://kling.ai/blog/kling-video-3-omni-native-lip-sync-audio-guide) says to keep the speaking face visible and to use short sentences.

6. **Street interview.** Suits Kling 3.0.
   > A reporter in a yellow raincoat holds a microphone on a wet street at dusk. Ambient rain and traffic. Medium shot. The reporter says calmly: "Tell me what you saw."
   First fix: if the lips lag, cut the line to six words or fewer.
7. **Chef to camera.** Suits Kling 3.0.
   > A chef in a white apron stands at a steel counter, chopping parsley. Medium close-up. The chef says warmly, smiling: "Fresh herbs change everything." Sound: knife on board, a pan sizzling.
   First fix: if the face turns away, add "looking straight at the camera".
8. **Two speakers.** Suits Kling 3.0.
   > Two friends sit at a cafe table, medium two-shot. Mia whispers, hesitant: "Did you tell him?" Leo answers in a low voice: "Not yet." Ambient cafe chatter.
   First fix: if the voices swap, add one distinguishing detail per speaker, such as "Mia in a blue scarf".
9. **Voiceover, no speaker on screen.** Suits Seedance 2.0.
   > A hiker climbs a ridge at sunrise, wide shot, handheld. A calm voiceover says, "Every summit starts as a bad idea." Wind, boots on gravel, no music.
   First fix: if the voice sounds scored or doubled, delete all style words from the sound line.
10. **Podcast host.** Suits Veo 3.1.
    > Medium close-up, static camera. A man in a grey sweater speaks into a studio microphone in a small recording room, warm lamp light. He says: "Today we cover one thing." SFX: soft chair creak. Ambient noise: room tone.
    First fix: if the sound is flat, write the room tone on its own "Ambient noise:" line.

## Short-form trend prompt examples

These follow common short-form formats. A viral format uses a repeatable structure, not a secret prompt. Pick the format, keep the prompt plain and make the subject your own. Seedance 2.0 suits them because fal's guide covers quoted lines and "no music".

11. **ASMR slice.** See [AI ASMR videos](https://honeyvid.ai/learn/ai-asmr-video/).
    > Macro, locked off. A knife slowly slices a glass-clear jelly cube on a wooden board. Soft diffuse light. Crisp knife contact, jelly wobble, no music.
    First fix: if the jelly does not cut, say "the blade passes fully through".
12. **Dancing pet.** See [AI cat dance videos](https://honeyvid.ai/learn/ai-cat-dance/).
    > A grey cat in tiny sunglasses stands upright on a kitchen counter and sways side to side in rhythm. Static medium shot, bright room. Soft paw taps, no music.
    First fix: if the legs bend wrong, say "stands on hind legs, forepaws raised".
13. **Fruit drama.** Original characters only.
    > Two cartoon-style strawberries argue on a kitchen table. The left strawberry says, "You forgot the cream." Close two-shot, bright sunlight. Quiet room tone, no music.
    First fix: if the faces look flat, add "round glossy fruit with simple eyes".
14. **Tiny world.** 
    > A miniature village built on a bread loaf, tiny people hanging washing. Macro tracking shot moves along the crust. Warm kitchen light. Soft ambient hum, no music.
    First fix: if the scale reads wrong, name one real object for scale, such as "a salt shaker".
15. **Satisfying loop.**
    > A glossy honey drop falls onto a stack of pancakes and spreads. Macro, slow push-in. Warm light. Liquid sound, no music.
    First fix: if the honey pools wrong, split into a pour shot and a spread shot.

## Camera move prompt examples

Camera moves read best when stated early. Google's Veo 3.1 guide puts cinematography first and gives a crane shot and a shallow depth of field example. The five prompts below are original.

16. **Crane up.** Suits Veo 3.1.
    > Crane shot starting low on a lone cyclist on a coastal road, rising to reveal the whole bay at sunrise. Soft morning light. Ambient noise: waves, a bike chain.
    First fix: if the rise is jerky, write one "slowly rising" and drop other motion words.
17. **Orbit.** Suits Veo 3.1.
    > Slow 90-degree orbit around a violinist on an empty stage, single spotlight, dust in the beam. SFX: a held violin note.
    First fix: if the player's face changes, shrink the orbit to 45 degrees.
18. **Push-in on a reaction.** Suits Veo 3.1.
    > Slow push-in to a close-up of a girl reading a letter at a window, rain on the glass. Cool blue tones. Ambient noise: steady rain.
    First fix: if the push-in is missing, put it in the first words of the prompt.
19. **Tracking walk.** Suits Veo 3.1.
    > Tracking shot behind a man walking through a night market, neon signs, steam from food stalls. Handheld feel. Ambient noise: crowd chatter, sizzling.
    First fix: if the crowd warps, ask for "sparse crowd".
20. **Rack focus.** Suits Veo 3.1.
    > Medium shot, rack focus from a key on a desk to a woman standing in the doorway, office at night, one desk lamp. Ambient noise: faint air conditioning.
    First fix: if focus never shifts, say which object is sharp at the start and the end.

## Image to video prompt examples

When you upload a picture, the picture carries the subject, so these prompts describe movement and sound only. Google's Veo 3.1 guide shows both start-and-end-frame and reference-image modes, and Kling lists reference elements built from images, video or voice.

21. **Portrait blink.** Suits Kling 3.0.
    > The woman slowly blinks, smiles and tilts her head slightly. Soft breeze moves her hair. Static camera.
    First fix: if the face drifts from the photo, ask for less movement and drop the head tilt.
22. **Food steam.** Suits Veo 3.1.
    > Steam rises from the bowl. A hand lifts a spoon. Slow push-in. SFX: spoon on ceramic.
    First fix: if the bowl changes, remove the spoon and keep the steam only.
23. **Start and end frame.** Suits Veo 3.1.
    > The camera performs a smooth arc from the front view to the view from behind, ending on the back of the seated man. Ambient noise: a quiet room.
    First fix: if the arc is wrong, describe the end frame in one sentence.
24. **Pet photo.** Suits Seedance 2.0.
    > The dog turns its head toward the camera and wags its tail. Soft indoor light, static medium shot. Tail thumps, no music.
    First fix: if the dog changes breed, cut to a single motion.
25. **Product still.** Suits Kling 3.0.
    > The camera slowly pushes in on the product. Light glints along the edge. Nothing else moves.
    First fix: if the product morphs, say "the product stays fixed and unchanged".

Use only photos you own or have rights to. Check each tool's face rules; see [Kling image to video](https://honeyvid.ai/learn/kling-ai-image-to-video/) for that model.

## Multi-shot story prompt examples

Veo 3.1 takes timed shots in one prompt. [Seedance 2.0's guide](https://fal.ai/learn/tools/seedance-2-0-prompting-guide) says to write cuts as "cut to" and calls three quick cuts in one take within range. Keep each shot to one action.

26. **Explorer, timed (Veo 3.1).** Adapted from the timed jungle sequence in Google's guide.
    > [00:00-00:03] Medium shot from behind a hiker pushing aside a vine to reveal a path. SFX: rustling leaves. [00:03-00:06] Reverse shot of the hiker's face, awed, ruins behind. [00:06-00:08] Wide high-angle crane shot of the hiker small before a temple half-covered in jungle.
    First fix: if shots merge, cut to two timed shots.
27. **Detective scene with references (Veo 3.1).** Adapted from the detective-and-office example in Google's guide.
    > Using the provided images for the detective and the office, create a medium shot of the detective behind his desk. He looks up and says in a weary voice, "Of all the offices in this town."
    First fix: if the room changes between shots, reuse the same office image for every shot.
28. **Coffee ad, three cuts (Seedance 2.0).**
    > Close-up of beans falling into a grinder. Cut to steam rising from a poured cup. Cut to a final wide shot of the cup on a table by a window, with a short title along the bottom. Room tone, no music.
    First fix: if the title is garbled, add the title in a later edit.
29. **Chase, three beats (Seedance 2.0).**
    > A courier runs through a crowded alley, handheld. Cut to a low angle as the courier vaults a crate. Cut to a wide shot of the courier reaching a door. Footsteps, breath, no music.
    First fix: if the vault fails, replace it with "jumps over a low box".
30. **Before and after (Seedance 2.0).**
    > A messy desk, static medium shot. Cut to the same desk, neat, same angle and light. Soft office hum, no music.
    First fix: if the desk differs in angle, say "same angle, same light" twice.

## How does this page differ from other prompt example pages?

Renderforest (7 Aug 2026) lists 55 templates and a separate prompt repair table. each::labs (12 Oct 2026) gives ten prompts with model suggestions and covers duration, framing and failure modes. This page differs in three ways: each prompt names the model whose documented conventions it follows, a suggested first fix sits directly beside every prompt, and the likeness and rights note below covers real people and characters.

## Can I use real people or characters in a viral AI video prompt?

Do not put a real person's name, face or voice, or a copyrighted character, in a prompt unless you have the right to. [YouTube requires creators to disclose](https://support.google.com/youtube/answer/14328491) realistic AI content, including a real person shown saying or doing something they did not.

Prompts 13 and 21 use original or owned subjects for that reason. Check the terms of the tool and platform you publish on before posting. For where Sora-era prompts stand now, see [Sora 2 prompts](https://honeyvid.ai/learn/sora-2-prompts/).

## Checklist before you run a prompt

Check each item once, then run a single clip. Costs add up per retry, so see the [AI video cost calculator](https://honeyvid.ai/learn/ai-video-cost-calculator/) before you plan a batch.

- One subject, one main action, one camera move.
- Dialogue written in your model's format (quotes for Seedance, speaker plus delivery for Kling, quotes plus SFX labels for Veo).
- "no music" written out if you do not want a score (fal's Seedance guide).
- A length your model allows (Kling 3 to 15 s, Seedance 4 to 15 s).
- No real person, celebrity or copyrighted character you lack rights to.
- Retry plan: change one element at a time.

## Questions

### How do I write a prompt for an AI video generator?

Write one subject, one action, one camera move, the light and the sound. Put the action first if you use Seedance. Keep it short: fal's Seedance guide says to spend your words on verbs, and Kling's guide asks you to pair acoustic details with the visual action.

### What is the best AI video prompt?

No single prompt is best, because Veo, Kling and Seedance each publish their own prompt conventions. The best prompt is the one written in your model's format, with one action and one camera move. Test it by changing only one element per retry.

### Can I copy a viral AI video prompt?

You can copy the structure, but check what the prompt asks for. Prompts that name a real person, a celebrity or a copyrighted character raise likeness and rights problems. YouTube also requires disclosure when realistic AI content shows a real person doing something they did not do.

### Do these prompts work in every model?

They are starting points written to one model's conventions each. When you switch models, check clip length (Kling 3 to 15 s, Seedance 4 to 15 s) and reference-input support, and adapt the sound and dialogue lines if the result needs it.

## Sources

- [The ultimate prompting guide for Veo 3.1](https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1), Google Cloud Blog (accessed 2026-10-26)
- [Seedance 2.0 prompting guide](https://fal.ai/learn/tools/seedance-2-0-prompting-guide), fal (accessed 2026-10-26)
- [Kling VIDEO 3.0 Omni: native lip sync and audio guide](https://kling.ai/blog/kling-video-3-omni-native-lip-sync-audio-guide), Kling AI (accessed 2026-10-26)
- [Disclosing use of altered or synthetic content](https://support.google.com/youtube/answer/14328491), YouTube Help (accessed 2026-10-26)
- [AI Video Prompt Examples: 50+ Templates That Work](https://www.renderforest.com/blog/ai-video-prompt-examples), Renderforest (accessed 2026-10-26)
- [AI video prompt examples that work](https://www.eachlabs.ai/blog/ai-video-prompt-examples-that-work), each::labs (accessed 2026-10-26)