Kling Image Generation API: Text-to-Image and References

Kling AI’s image generation endpoint creates high-quality images from text descriptions, optionally guided by a reference image for character consistency or face preservation. All requests are asynchronous: submit a task and poll for results using the returned task_id. The request format is fully compatible with the official Kling API.

export TOKEN="oh-xxxxxxxxxxxxxxxx"

Authorization: Bearer <TOKEN>
Content-Type: application/json

Endpoints

Action	Method	Path
Generate images	`POST`	`/kling/v1/images/generations`
Multi-image reference	`POST`	`/kling/v1/images/multi-image2image`
Omni image (o1)	`POST`	`/kling/v1/images/omni-image`
Query task	`GET`	`/kling/v1/images/generations/{task_id}`

Request Parameters

Parameter	Type	Required	Description
`model_name`	string	No	Model to use (default: `kling-v1`). See models below.
`prompt`	string	Yes	Text description of the image. Max 2500 characters.
`negative_prompt`	string	No	Elements to exclude. Not supported in image-to-image mode (when `image` is set).
`image`	string	No	Reference image — HTTPS URL or base64. Enables image-to-image mode.
`image_reference`	string	No	Reference type: `subject` (character features) or `face` (facial appearance). Required when `image` is set and `kling-v1-5` is used.
`image_fidelity`	number	No	How closely to follow the reference image `[0, 1]`.
`human_fidelity`	number	No	Face similarity strength when `image_reference: face[0, 1]`.
`resolution`	string	No	Output resolution: `1k` (standard) or `2k` (high-definition). Default: `1k`.
`n`	integer	No	Number of images to generate `[1, 9]`. Default: 1.
`aspect_ratio`	string	No	Aspect ratio — see options below. Default: `16:9`.
`watermark_info`	object	No	Watermark settings.
`callback_url`	string	No	Webhook URL for completion notifications.
`external_task_id`	string	No	Your own task ID for idempotency and tracking.

Aspect Ratio Options

16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9

Supported Models

Model	Capabilities
`kling-v1`	Text-to-image, image-to-image
`kling-v1-5`	Text-to-image, image-to-image with `subject`/`face` reference
`kling-v2`	Text-to-image, multi-image reference
`kling-v2-new`	Style transfer (image-to-image)
`kling-v2-1`	Text-to-image, multi-image reference
`kling-image-o1`	Omni image generation

When using kling-v1-5 with image set, you must also set image_reference to either subject or face. Other models do not support image_reference.

Examples

Text-to-Image

Generate a single high-resolution image from a text prompt:

curl --request POST \
  --url https://aiapi.fhddos.com/kling/v1/images/generations \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
    "model_name": "kling-v1",
    "prompt": "A beautiful sunset over the ocean, vibrant colors, photorealistic, high detail",
    "resolution": "2k",
    "n": 1,
    "aspect_ratio": "16:9"
  }'

Character Reference (Subject)

Use a reference image to preserve a character’s visual appearance:

curl --request POST \
  --url https://aiapi.fhddos.com/kling/v1/images/generations \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
    "model_name": "kling-v1-5",
    "prompt": "The character holding a glowing magic wand in an enchanted forest",
    "image": "https://example.com/character-reference.jpg",
    "image_reference": "subject",
    "image_fidelity": 0.8,
    "resolution": "2k",
    "aspect_ratio": "3:4"
  }'

Face Reference

Generate a portrait that preserves facial features from a reference photo:

curl --request POST \
  --url https://aiapi.fhddos.com/kling/v1/images/generations \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
    "model_name": "kling-v1-5",
    "prompt": "A professional business headshot portrait, studio lighting, clean background",
    "image": "https://example.com/face-photo.jpg",
    "image_reference": "face",
    "human_fidelity": 0.9,
    "resolution": "2k",
    "aspect_ratio": "1:1"
  }'

Batch Generation

Generate multiple image variations from a single prompt:

curl --request POST \
  --url https://aiapi.fhddos.com/kling/v1/images/generations \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
    "model_name": "kling-v2",
    "prompt": "Abstract colorful digital artwork, modern minimalist style",
    "n": 4,
    "resolution": "1k",
    "aspect_ratio": "1:1"
  }'

Style Transfer (kling-v2-new)

Apply the visual style of a reference image to a new composition:

curl --request POST \
  --url https://aiapi.fhddos.com/kling/v1/images/generations \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
    "model_name": "kling-v2-new",
    "prompt": "A futuristic cityscape at night with neon reflections",
    "image": "https://example.com/style-reference.jpg",
    "resolution": "2k",
    "aspect_ratio": "16:9"
  }'

Omni Image (kling-image-o1)

Use the Omni image model for complex, instruction-following generation:

curl --request POST \
  --url https://aiapi.fhddos.com/kling/v1/images/omni-image \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
    "model_name": "kling-image-o1",
    "prompt": "A vintage bicycle leaning against a cobblestone wall in a European alley, afternoon sunlight, film grain",
    "resolution": "2k",
    "aspect_ratio": "3:2"
  }'

Responses

Task Created

{
  "code": 0,
  "message": "success",
  "request_id": "req_1735558800_img123",
  "data": {
    "task_id": "task_01JGHP...",
    "task_status": "submitted",
    "created_at": 1735558800000,
    "updated_at": 1735558800000
  }
}

Task Succeeded

When you poll and task_status is succeed:

{
  "code": 0,
  "message": "success",
  "data": {
    "task_id": "task_01JGHP...",
    "task_status": "succeed",
    "task_result": {
      "images": [
        {"index": 0, "url": "https://cdn.example.com/image-0.jpg"},
        {"index": 1, "url": "https://cdn.example.com/image-1.jpg"}
      ]
    }
  }
}

Polling for Results

Query a task using either task_id or the Fhddos platform_id:

# Single task
curl "https://aiapi.fhddos.com/kling/v1/images/generations/$TASK_ID" \
  -H "Authorization: Bearer $TOKEN"

# Task list
curl "https://aiapi.fhddos.com/kling/v1/images/generations?pageNum=1&pageSize=30" \
  -H "Authorization: Bearer $TOKEN"

For the full response schema across all task states, see Task Query.

Multi-Image Reference

To fuse elements from multiple reference images, use the dedicated endpoint:

curl --request POST \
  --url https://aiapi.fhddos.com/kling/v1/images/multi-image2image \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{
    "model_name": "kling-v2-1",
    "prompt": "Combine these visual styles into a cohesive product photo",
    "images": [
      "https://example.com/ref1.jpg",
      "https://example.com/ref2.jpg"
    ],
    "resolution": "2k",
    "aspect_ratio": "1:1"
  }'

Refer to the official Kling documentation for the exact field schema for multi-image2image.

Constraints Summary

prompt must be 2500 characters or fewer
negative_prompt is not supported in image-to-image mode (when image is provided)
image_reference is required when using kling-v1-5 with an image — and is not supported on other models
n range is 1–9

​Endpoints

​Request Parameters

​Aspect Ratio Options

​Supported Models

​Examples

​Text-to-Image

​Character Reference (Subject)

​Face Reference

​Batch Generation

​Style Transfer (kling-v2-new)

​Omni Image (kling-image-o1)

​Responses

​Task Created

​Task Succeeded

​Polling for Results

​Multi-Image Reference

​Constraints Summary

Endpoints

Request Parameters

Aspect Ratio Options

Supported Models

Examples

Text-to-Image

Character Reference (Subject)

Face Reference

Batch Generation

Style Transfer (kling-v2-new)

Omni Image (kling-image-o1)

Responses

Task Created

Task Succeeded

Polling for Results

Multi-Image Reference

Constraints Summary