Prompt Kling for image-to-video
Last updated: October 12, 2026
Kling turns an approved still into a short clip that follows the camera and the action you write, beat by beat. Pick it when the frame is signed off and you need a specific move: a spell cast, an attack wind-up, an idle with a slow camera push.
Let the agent write it
You don't have to write any of this yourself. The Layer agent knows how to prompt each model it runs, from the same guidance this page is built on. Attach the still in chat, name the model and say what should happen, for example: Animate this with Kling V3 Standard: the mage raises his staff and fires a fireball to the right, fixed camera, 3 seconds. The agent writes the prompt, shows you its plan, and runs it when you approve. Use the rest of this guide when you want to write or adjust the prompt yourself, and see Write a brief for the agent for how to brief it.
When to use it
Use Kling when the motion has a clear order: wind-up, release, follow-through. It plays written beats in sequence and keeps the character's design, colours and style from the first frame, which suits ability casts, attack and emote animations, and short trailer shots. Kling V3 Standard also splits one clip into several shots and generates sound in the same pass.
It loses the character in acrobatics: leaps, spins and anything that bounces or flies in a complex arc. Cut that into separate clips. For slow ambient motion with native audio, compare Prompt Seedance for image-to-video. To generate a clip with no image at all, see Create a video.
Versions on Layer
Kling V3 Standard
The version this guide's examples were made with. Clips run 3 to 15 seconds, with a first and a last frame, native audio, a negative prompt, and up to six shots in one clip. It follows the order of written beats closely.
Kling v2.6 Pro
Clips of 5 or 10 seconds with a first and a last frame, a negative prompt and native audio. One shot per clip.
Kling v2.5 Turbo Pro
Clips of 5 or 10 seconds with a first and a last frame, and no sound. In our test it kept the character on-model at full resolution but took its own route to the end of the action: given the same spell-cast prompt, it fired to the left before turning to fire to the right. Write the beats even more plainly, or use Kling V3 Standard when the order matters.
How to write the prompt
Camera, then the action in short sentences, one beat each, then the style:
Fixed camera, medium shot. The mage slowly raises the staff overhead with both hands. The ember on the staff swells into a ball of fire. He thrusts the staff forward and the fireball flies out of frame to the right. His cloak billows back. Stylised painterly game art, colours unchanged.
Kling's own formula is subject, movement, scene, camera, lighting and atmosphere. The first frame already fixes the subject, the scene and the light, so the prompt goes on movement and camera. Short sentences give the model one instruction at a time, and it plays them in the order you write them. That prompt is the one in the clip above: every beat lands in 3 seconds.
What works
Short sentences, slow action
Write one movement per sentence, at a speed a body can manage: "The mage slowly raises the staff overhead with both hands. The ember on the staff swells into a ball of fire." We ran this against "The mage explodes into action, leaps high into the air, spins three times and slams the staff into the ground, unleashing a violent fiery explosion that sends rocks bouncing in every direction" on the same frame and seed. The short-sentence prompt held the camera and kept the mage on-model through every beat. The burst prompt turned him away from the camera, threw him into the air with his face lost, and ended in an explosion that was no use as a game asset. Kling's guide says why: counts and complex physics, such as bouncing objects and thrown arcs, are what it handles worst.
Name the camera
Open with the camera: fixed camera, slow push-in, tracking shot, low-angle shot. In our tests, "Slow push-in toward the mage" gave a steady, gentle move in while the cloak and ember animated, and "Fixed camera" held the frame still through the whole cast. Leave the camera out and you get a static or slightly drifting shot.
Motion, not description
The start image is the anchor: it fixes the character, the layout and any lettering in it. Spend the prompt on what changes from that frame, and leave out what the image already shows.
Name the style
End with the medium ("stylised painterly game art", "2D anime", "stylised 3D game cinematic") and "colours unchanged", so stylised art keeps its look while it moves.
Several shots in one clip
On Kling V3 Standard, the prompt box can split one clip into up to six timed shots. Write each shot's camera and action on its own, and describe the character the same way in every shot so it stays the same person.
Sound and dialogue
With Generate audio on (Kling V3 Standard and Kling v2.6 Pro), write the action first, then the line, tagged with its speaker and tone: Mage (low, steady voice): "Burn." For a second line straight after the first, start it with "Immediately" so the two do not merge.
What to avoid
Acrobatics and complex physics
Leaps, spins, bouncing debris and counted objects ("three fireballs") come out wrong or off-model. Keep each clip to a few slow, continuous movements, and cut a fight into several clips.
Long lists of exclusions
Kling V3 Standard, Kling v2.6 Pro and Kling v2.5 Turbo Pro have a Negative prompt field. Give it a few specific artefacts (blur, distortion, extra limbs) rather than a long list.
Asking for new text
Kling keeps lettering that is already in the start image readable as the scene moves, but do not ask it to write new titles or logos. Paint them into the frame first, or add them afterwards in the video editor.
Settings that matter
First frame and Last frame pin where the clip starts and ends. Duration starts at 3 seconds on Kling V3 Standard and 5 seconds on the 2.x versions. Generate audio adds sound in the same pass on Kling V3 Standard and Kling v2.6 Pro. Aspect ratio sets the shape and size of the clip. Under the advanced settings, Negative prompt takes what to keep out, and Guidance scale sets how closely the result follows the prompt.
Prompts to start from
Spell cast
Fixed camera, medium shot. The mage slowly raises the staff overhead with both hands. The ember on the staff swells into a ball of fire. He thrusts the staff forward and the fireball flies out of frame to the right. His cloak billows back. Stylised painterly game art, colours unchanged.
Character idle with a camera move
Slow push-in toward the mage. He breathes slowly, his cloak and hood stirring in a light breeze. The ember on his staff flickers and sheds small sparks that drift upward. Stylised painterly game art, colours unchanged.
Use it in Layer
Add the approved still to a session's board.
Click View on the image, then Turn into video. Create video opens in the side panel with the image as its First frame.
Click the model name at the top of the panel and choose Kling V3 Standard. The first frame stays in place.
In the prompt box, write the camera, then the action one short sentence at a time, then the style.
Click Create video. The cost shows on the button before you click.
When the clip lands on the board, click View to open it in the viewer and play it. Then change one thing in the prompt at a time and run it again.
Learn more
- Kling VIDEO 3.0 model user guide (Kling AI): what 3.0 adds (3 to 15 second clips, multi-shot, native audio and dialogue) and how to name shots and speakers.
- Text-to-video prompt guide (Kling AI): the prompt formula, simple sentences, and why counts and complex physics go wrong.
- Kling 3.0 prompting guide (fal): writing in shots rather than descriptions, explicit camera behaviour, and treating the input image as the anchor the motion starts from.
- Kling Video v3 Standard image to video (fal): the image-to-video inputs with a worked camera-and-motion prompt.