Kling AI Prompts: 9 Ready-to-Use Templates & Test Results

👁 125 reads

A good Kling prompt describes motion, not the image: who moves, how the camera travels, what happens in the background, and what should stay still. Below you’ll find the structure we use, nine ready-to-use English prompts with descriptions, and lessons from five tests where Kling received exactly these instructions. For versions, pricing, and competitor comparisons, see our Kling guide.

The Prompt Structure We Use

We arrange the prompt in a fixed order, one sentence per element. This way, when something doesn’t work, we know which sentence to fix.

1. Shot and camera - “Slow push-in”, “static camera on a tripod”, “handheld tracking shot from behind”.
2. Main motion - what the character or object does, using a verb: “pours”, “turns”, “walks to the window”.
3. Background motion - background, weather, pedestrians, leaves.
4. What should stay still - “The cup and spoon stay still”, “keep the label readable”.
5. Light and style - “soft overcast daylight”, “realistic”, “35mm film look”.
6. Sound and exclusions - “quiet cafe ambience, no music, no text, no subtitles”.

For image-to-video, we skip describing the scene’s appearance because the photo establishes it. Describing what’s already visible can prompt the model to “correct” the frame.

Ready-to-Use Kling Prompts

Replace square brackets with your own content. For each prompt, we note whether we tested it ourselves.

1. Image-to-Video: Animating a Product Photo

“Slow camera push-in toward the [product]. [One small natural movement, e.g. steam rises from the cup]. The [product] stays perfectly still and keeps its shape and label. Soft natural light, realistic, no text, no music.”

Tested October 3, 2026 on Kling 3.0 Pro (coffee version, results described below).

2. Text-to-Video: Food or Beverage Scene

“Close-up of hot coffee being poured into a white ceramic cup on a wooden kitchen table, steam rising, morning sunlight through the window, shallow depth of field, slow gentle camera push-in, realistic.”

Tested September 27, 2026 on Kling 3.0 Standard: sharp porcelain and steam, but the coffee level barely rises over 4 seconds.

3. Text-to-Video: Street and City

“A [colour] tram passing along a rainy street at dusk, reflections of street lights on wet cobblestones, people with umbrellas on the sidewalk, static camera on a tripod, cinematic realistic footage.”

Tested September 27, 2026 on Kling 3.0 Standard: the tram came out cream-red instead of yellow-red, and the light resembled afternoon, not dusk.

4. Character from Reference in New Scene (Omni)

“@Image1 is the main character, keep her face, hair and clothes exactly the same. She [action] in [place]. [Camera move], [light]. She turns to the camera and says in English: “[line]”. No subtitles, no music.”

Tested September 28, 2026 on Kling 3.0 Omni Standard in a version without “turns to the camera”: the character matched, but spoke the line with her back to the camera. That’s why we add the turn before speech.

5. Character Speaking to Camera

“Medium close-up of [person] sitting at [place], looking straight into the camera. [He/She] says clearly and calmly in English: “[line]”. Natural daylight from a window, static camera, realistic, no subtitles, no on-screen text, no music.”

Tested September 28, 2026 on Kling 3.0 Turbo and Pro: English sentence flawless, Polish with foreign pronunciation. Details in the article Kling po polsku.

6. Product Ad with Camera Movement

“Slow 180-degree orbit around the [product] on a [surface]. Studio lighting with a soft rim light. The label text stays sharp and readable. Clean background, no people, no extra objects, no text overlays.”

Not tested.

7. Multiple Shots in One Clip (Kling 3.0 and Newer)

“Shot 1 (0-4 s): wide shot of [place], [action]. Shot 2 (4-8 s): close-up of [detail], [action]. Shot 3 (8-12 s): [character] looks at the camera and smiles. Same character and outfit in every shot, cut on action.”

Not tested. In the API, this is done with the multi_prompt parameter with a separate prompt for each shot.

8. Motion from Reference Video (Omni and Motion Control)

“@Element1 performs the exact movement from @Video1. Keep the camera angle and timing of @Video1, replace the background with [place].”

Not tested. This template comes from our article about Kling 4.0 Omni Reference.

9. Negative Prompt

In the Kling 3.0 API, the default negative prompt is “blur, distort, and low quality”. We add to it whatever ruins our specific scene, for example “extra fingers, warped text, flickering, morphing objects”. In the app, the negative field is in advanced settings.

Our Test: Cafe Prompt on Kling 3.0 Pro

On October 3, 2026, we gave Kling 3.0 Pro a photo of a coffee cup in a Krakow cafe and prompt number 1 in full: camera push-in, steam over coffee, blurred pedestrians behind the window, leaves in the wind, cup stays still, no music or text.

Conclusion: Kling follows the first sentence (camera) and prohibitions best, but loses subtle physical effects from later in the prompt. When steam is most important to us, we move it to the beginning and add timing: “Thick steam rises continuously from the coffee during the whole shot”. MiniMax H3 Max with the same prompt produced steam and leaves without corrections - the comparison is in the Kling guide.

Five Rules from Our Tests

• Most important first. Elements from the end of long prompts were executed least often by Kling.
• Colors and time of day explicitly. “Yellow and red tram” produced a cream-red car. For important colors, we write them twice or provide an input image.
• Change over time requires continuous verbs. “Coffee being poured” wasn’t enough for the coffee level to rise. Better: “the cup fills up from empty to full”.
• Spoken lines in English. Kling reads Polish sentences with foreign phonetics.
• Character direction before speech. Without “turns to the camera”, the character may speak with their back turned.

Frequently Asked Questions

Should You Write Kling Prompts in Polish?

Better in English. Kling understands Polish prompts, but follows camera movement details worse, and speaks Polish lines with a foreign accent.

How Long Should a Kling Prompt Be?

4-6 sentences work best for us, one per scene element. The Kling 3.0 API accepts up to 2,500 characters, Kling 4.0 according to the manufacturer up to 8,000 tokens, but a longer prompt doesn’t mean better results.

Can You Copy Prompts from Midjourney to Kling?

Only as appearance descriptions for input images. For Kling, you need to add motion: camera, action, and background, because a static description produces a static clip.

Where Do You Enter a Negative Prompt in Kling?

In the app in advanced task settings, and in the API in the negative_prompt parameter. By default, the API enters “blur, distort, and low quality”.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Sources and Verification Date

We conducted tests on September 27 and 28 and October 3, 2026, via the fal.ai and OpenRouter APIs, each clip in one attempt.

• fal.ai - Kling 3.0 Pro, API parameters
• OpenRouter - Kling Video v3.0 Pro
• Kling AI - Kling 4.0 vs 3.0