live
MODEL IDklingai:kling-image@4.0

Kling IMAGE 4.0

Kling AI
by

Kling IMAGE 4.0 is an image model for text-to-image and reference-guided generation at native 2K, 4K, and 8K, with natural lighting and skin and material detail close to a camera photo. It renders up to 2,000 characters of legible text in seven languages, combines up to 10 reference images and elements, and plans series of up to 10 images that keep subjects and style consistent. It also designs e-commerce product pages and websites as a single tall image, which suits storyboards, slide decks, product shoots, and campaign visuals.

Kling IMAGE 4.0

Prompting

How to write text-to-image prompts for Kling IMAGE 4.0: picking a tier or an exact size, deciding when 8K pays off, describing people, and pinning HEX colors.

Introduction

Kling IMAGE 4.0 is built around images that look photographed: soft light falloff and skin with visible pores. Write the prompt the way you would brief a photographer, with the people, what they wear and do, the place, the light and the camera each in a clause of its own.

Group shots keep every face distinct. The prompt gave each of the six runners an age, a build and a hairstyle of their own, and "each face has its own shape, skin tone and expression" asks the model not to repeat a face across the row.

This guide covers the request and its two ways of setting the size, when 8K is worth it, how to describe people and the relationship between two of them, and how to pin exact colors. Text, edits and references have their own guides: text and typography, editing images and reference images.

The request

A text-to-image request needs only a prompt. This is the request behind the hero:

Try in Playground
import { createClient } from '@runware/sdk'

const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()

const [result] = await client.run({
  model: 'klingai:kling-image@4.0',
  positivePrompt: 'A candid phone photo for a running club\'s social feed, taken right after a Sunday morning 10K. Six club members stand close together in a loose row on a riverside path, sweaty and laughing, race bibs still pinned to their shirts. A tall man in his sixties with a white buzz cut and a gray beard, a short woman in her twenties with box braids tied up in a high bun, a broad-shouldered man in his forties with a shaved head and a sleeve tattoo, a freckled teenage girl with a red ponytail, a slim woman in her fifties with silver-streaked curls and reading glasses pushed up on her head, and a stocky man in his thirties with a thick black mustache holding a water bottle. Each face has its own shape, skin tone and expression. Soft overcast morning light, flat even shadows, a few joggers blurred in the background. Shot at eye level on a phone at arm\'s length, slightly off-center framing, natural skin texture with pores and flushed cheeks, true-to-life color, no text, no logos.',
  width: 2496,
  height: 1664
})
import asyncio
import os

from runware import Runware


async def main():
    async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
        results = await client.run({
            "model": "klingai:kling-image@4.0",
            "positivePrompt": "A candid phone photo for a running club's social feed, taken right after a Sunday morning 10K. Six club members stand close together in a loose row on a riverside path, sweaty and laughing, race bibs still pinned to their shirts. A tall man in his sixties with a white buzz cut and a gray beard, a short woman in her twenties with box braids tied up in a high bun, a broad-shouldered man in his forties with a shaved head and a sleeve tattoo, a freckled teenage girl with a red ponytail, a slim woman in her fifties with silver-streaked curls and reading glasses pushed up on her head, and a stocky man in his thirties with a thick black mustache holding a water bottle. Each face has its own shape, skin tone and expression. Soft overcast morning light, flat even shadows, a few joggers blurred in the background. Shot at eye level on a phone at arm's length, slightly off-center framing, natural skin texture with pores and flushed cheeks, true-to-life color, no text, no logos.",
            "width": 2496,
            "height": 1664
        })


asyncio.run(main())
curl https://api.runware.ai/v1 \
  -H "Authorization: Bearer $RUNWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '[
    {
      "taskType": "imageInference",
      "taskUUID": "7b3e9c41-2d8a-4f65-b1e7-9c0a4d2f6e83",
      "model": "klingai:kling-image@4.0",
      "positivePrompt": "A candid phone photo for a running club's social feed, taken right after a Sunday morning 10K. Six club members stand close together in a loose row on a riverside path, sweaty and laughing, race bibs still pinned to their shirts. A tall man in his sixties with a white buzz cut and a gray beard, a short woman in her twenties with box braids tied up in a high bun, a broad-shouldered man in his forties with a shaved head and a sleeve tattoo, a freckled teenage girl with a red ponytail, a slim woman in her fifties with silver-streaked curls and reading glasses pushed up on her head, and a stocky man in his thirties with a thick black mustache holding a water bottle. Each face has its own shape, skin tone and expression. Soft overcast morning light, flat even shadows, a few joggers blurred in the background. Shot at eye level on a phone at arm's length, slightly off-center framing, natural skin texture with pores and flushed cheeks, true-to-life color, no text, no logos.",
      "width": 2496,
      "height": 1664
    }
  ]'
runware run klingai:kling-image@4.0 \
  positivePrompt="A candid phone photo for a running club's social feed, taken right after a Sunday morning 10K. Six club members stand close together in a loose row on a riverside path, sweaty and laughing, race bibs still pinned to their shirts. A tall man in his sixties with a white buzz cut and a gray beard, a short woman in her twenties with box braids tied up in a high bun, a broad-shouldered man in his forties with a shaved head and a sleeve tattoo, a freckled teenage girl with a red ponytail, a slim woman in her fifties with silver-streaked curls and reading glasses pushed up on her head, and a stocky man in his thirties with a thick black mustache holding a water bottle. Each face has its own shape, skin tone and expression. Soft overcast morning light, flat even shadows, a few joggers blurred in the background. Shot at eye level on a phone at arm's length, slightly off-center framing, natural skin texture with pores and flushed cheeks, true-to-life color, no text, no logos." \
  width=2496 \
  height=1664
{
  "taskType": "imageInference",
  "taskUUID": "7b3e9c41-2d8a-4f65-b1e7-9c0a4d2f6e83",
  "model": "klingai:kling-image@4.0",
  "positivePrompt": "A candid phone photo for a running club's social feed, taken right after a Sunday morning 10K. Six club members stand close together in a loose row on a riverside path, sweaty and laughing, race bibs still pinned to their shirts. A tall man in his sixties with a white buzz cut and a gray beard, a short woman in her twenties with box braids tied up in a high bun, a broad-shouldered man in his forties with a shaved head and a sleeve tattoo, a freckled teenage girl with a red ponytail, a slim woman in her fifties with silver-streaked curls and reading glasses pushed up on her head, and a stocky man in his thirties with a thick black mustache holding a water bottle. Each face has its own shape, skin tone and expression. Soft overcast morning light, flat even shadows, a few joggers blurred in the background. Shot at eye level on a phone at arm's length, slightly off-center framing, natural skin texture with pores and flushed cheeks, true-to-life color, no text, no logos.",
  "width": 2496,
  "height": 1664
}
Response
[
  {
    "taskType": "imageInference",
    "taskUUID": "7b3e9c41-2d8a-4f65-b1e7-9c0a4d2f6e83",
    "imageUUID": "e14a7c2d-9b35-4f08-a6d1-3c8e5f2b7a90",
    "imageURL": "https://im.runware.ai/image/os/a14d18/ws/2/ii/e14a7c2d-9b35-4f08-a6d1-3c8e5f2b7a90.jpg"
  }
]
  • positivePrompt can run long. An English prompt takes up to 50,000 characters and a prompt in a non-Latin script up to 21,000, and a mix of the two is weighted between those limits.
  • The size comes from resolution or from width and height, never both. Leave both out and you get a 2K image in a ratio the model picks.
  • resolution is a tier: 2K, 4K or 8K.
  • width and height take one of eight aspect ratios at each tier, from 21:9 to 9:16. The exact pairs are listed in the API reference.

Choosing the size

Resolution tiers

A tier sets the pixel count and the model picks the shape from the prompt. The meal-kit ad below sent resolution: "2K" and no dimensions, so its prompt says "vertical 9:16" in words.

Name the orientation whenever you send a tier. Without it the prompt is the only hint the model has about the shape, and a scene described without one can come back in any of the eight ratios.

Exact sizes

Send width and height when the frame is fixed in advance, such as a 3:2 feed post or a 9:16 story slot. The hero sent 2496 × 1664, which is the 2K pair for 3:2.

When to ask for 8K

8K is for images that will be cropped or printed large. At page width an 8K render looks like a 2K one, and the extra pixels only show once you zoom in. The harbor below was rendered at 10880 × 6144 for a billboard:

The crop is one eighth of the width and one eighth of the height of the original, and it still holds the shutters and a laundry line as separate objects, down to the single roof tiles. That is the margin a billboard or a cropped hero needs. For a web image shown at its own size, 2K already has every pixel the screen can display.

Describing people

Kling recommends a fixed order for a portrait prompt: the person, then clothing and accessories, then expression and action, then the scene. Lighting and texture close it. The skincare portrait follows that order:

A woman in her late twenties with deep brown skin and a short natural afro, a light dusting of freckles across her nose, wearing an oversized ivory linen shirt and small gold huggie earrings, She sits sideways on a window ledge with her knees pulled up, her head tilted back against the frame, eyes half closed and a small relaxed smile, A bright bathroom with white tiles and a trailing pothos plant behind her, softly out of focus, Soft morning light through a frosted window, gentle highlights on her cheekbones and forehead, preserved shadow detail, Medium close-up at eye level, 85mm lens, shallow depth of field, Realistic skin texture with visible pores, subtle film grain, no makeup look
PersonWardrobeExpression and actionSceneLightingCameraTexture

Expression carries the mood. "Head tilted back, eyes half closed" reads as release, and the same woman looking away with her lips parted would read as distant. Describe the pose and the face, and let the viewer infer the feeling from them.

Light sets the register. Soft window light and preserved shadows suit a calm brand. Direct flash with hard shadows gives a playful or rebellious one, and a low backlit sunset a more restrained one.

Two people in one frame

With two people, describe how they relate: where each one stands, how far apart they are and where each one looks. The co-founders below are placed and directed separately, and the line "glancing up at her and laughing at something she just said" ties them together.

Give each person a different gaze. One looking at the camera and one looking at the other makes a story out of a group shot. Two people both facing the lens read as a posed photo.

Exact colors

Write a color as a HEX code next to the object it belongs to. "A cropped track jacket in #FF6B35" pins the jacket to the brand's orange, where "an orange jacket" leaves the shade to the model.

The prompt gives one code per surface, the jacket, the tights, the wall and the stripe, so each value has exactly one place to go. Flat, matte surfaces hold their code best: the wall came back as #254BF6 against #1F4FFF, while highlights and shadows spread the glossy jacket from #DD6032 to #F68951 around #FF6B35.

Tips

  1. Name the orientation when you send a tier. With resolution alone, "vertical 9:16" or "wide 21:9" in the prompt decides the shape.

  2. Save 8K for crops and print. At screen size it looks like 2K, and the extra detail only shows once you zoom in.

  3. Give every person in a group their own traits. Age, build and hair per person keep faces from repeating across the frame.

  4. Describe expression as pose and face. "Eyes half closed, head tilted back" gives the model something to render, where "relaxed" does not.

  5. Direct two people separately, then link them. Position, distance and gaze for each one, and one line about what passes between them.

  6. Put HEX codes on named surfaces. One code per object, and flat, matte surfaces render closest to it.