페이지 이동경로
  • Docs>
  • Karlo>
  • REST API

Karlo

REST API

This document describes how to integrate Karlo APIs into your service with a REST API.

Version upgrade

The Karlo model version for Karlo APIs has been changed to 2.0 on July 6, 2023, and to 2.0.4.0 on November 10, 2023 with additional features related to image editing. The new version offers enhanced features than previous versions.

Before you begin

Image file specification

The original image must comply with the specifications below.

  • File size: 2 MB or less
  • Image size: 2048 pixels or less for width and height (Recommended)
  • Ratio: 1:1 (Recommended)
    • Images of other ratios will be resized to a 1:1 ratio.
  • File format: JPG, PNG, webp (Recommended)

Image encoding and decoding

Karlo API handles the image file for the request and the response as a Base64 encoded string value.

  • Request: Pass a Base64 encoded string value of the image file through the image parameter.
  • Response: Base64 decode the value of the image field and save it as a PNG image file.

Below is a sample for Base64 encoding and decoding.

import io
import base64
from PIL import Image

# Base64 encoding
def imageToString(img):
    img_byte_arr = io.BytesIO()
    img.save(img_byte_arr, format='PNG')
    my_encoded_img = base64.encodebytes(img_byte_arr.getvalue()).decode('ascii')
    return my_encoded_img

# Base64 decoding and saving as a PNG file
def stringToImage(base64_string, mode='RGBA'):
    imgdata = base64.b64decode(str(base64_string))
    img = Image.open(io.BytesIO(imgdata)).convert(mode)
    return img

# Image loading
img = Image.open('my_image.png')

# Base64 encoding
img_base64 = imageToString(img)

# Decoding and saving
image = stringToImage(img_base64, mode='RGB')

Generate image

Basic information
Method URL Authorization
POST https://api.kakaobrain.com/v2/inference/karlo/t2i REST API Key
Permission Prerequisite Kakao Login User consent
- - - -

Create new images based on the given text prompt.

Include the REST API key in the request header, and send a POST request. You must specify the prompt in the request body. Refer to How to use to request more efficiently with optional parameters.

If failed, refer to Error code to figure out the reason.

Request

Header
Name Description Required
Authorization Authorization: KakaoAK ${REST_API_KEY}
REST API key as a type of user authentication.		 O
Content-Type Content-Type: application/json
The data type of the request.		 O
Body
Name Type Description Required
prompt String A text describing the desired image.
(Maximum: 2048 characters)		 O
negative_prompt String A text describing anything you want to exclude or control from the image when it is generated.
(Maximum: 2048 characters)		 X
width Integer Width of the image, a multiple of 8.
(Unit: pixel, Default: 512, Minumum: 384, Maximum: 640)		 X
height Integer Height of the image, a multiple of 8.
(Unit: pixel, Default: 512, Minumum: 384, Maximum: 640)		 X
upscale Boolean Whether to upscale the generated image.
true: Upscale
false: Do not upscale		 X
scale Integer Rate to upscale the image, 2 or 4.
(Default: 2)

Note: Even if the scale is set to 4, you can only upscale the image up to 2048 pixels in width and height.		 X
image_format String Format to save the image in. One of:
webp
jpeg
png
(Default: webp)		 X
image_quality Integer Save Image Quality.
(Default: 70, Minumum: 1, Maximum: 100)		 X
samples Integer Number of images to generate.
(Default: 1, Minumum: 1, Maximum: 8)		 X
return_type String Type of the image to return. One of:
base64_string: A Base64 encoded value of the image file
url: A URL of the image file
(Default: url)		 X
prior_num_inference_steps Integer The number of steps of the prior denoising process.
(Default: 25, Minumum: 10, Maximum: 100)		 X
prior_guidance_scale Double Guidance scale of the prior denoising process.
(Default: 5.0, Minumum: 1.0, Maximum: 20.0)		 X
num_inference_steps Integer The number of steps of the decoder denoising process.
(Default: 50, Minumum: 10, Maximum: 100)		 X
guidance_scale Double Guidance scale of denoising process.
(Default: 5.0, Minumum: 1.0, Maximum: 20.0)		 X
scheduler String Scheduler used by decoder denoising process, One of:
decoder_ddim_v_prediction
decoder_ddpm_v_prediction
(Default: decoder_ddim_v_prediction)		 X
seed Integer[] A seed of length equal to the batch size length.
seed must have the same lengh with the number of samples.
Composed of numbers greater than or equal to 0 and less than or equal to 4,294,967,295.
If none, a random.
(Default: null)

Note: Karlo always generates the same image, if all parameters, including seed, have the same value.		 X
nsfw_checker Boolean Whether to perform Detect NSFW content on the generated image.
true: Perform
false Do not perform
(Default: false)		 X
face_refiner FaceRefiner Setting parameters for Refine facial structure. X
FaceRefiner
Name Type Description Required
bbox_size_threshold Double Ratio of facial area to total image to apply face shape refinement feature, apply feature only if actual facial area is smaller than set size
(Default: 1, Minumum: 0, Maximum: 1)		 X
bbox_filter_threshold Double Threshold to determine if the image is a human face, not applied if 0.
(Default: 1, Minumum: 0, Maximum: 1)		 X
restoration_repeats Double Number of facial structure refinements applied
(Default: 2, Minumum: 1, Maximum: 10)		 X
weight_sft Double Original image retention weight determines how much of the original image is retained.
The higher the value, the more of the original image is preserved.
(Default: 0.5, Minumum: 0, Maximum 1)		 X

Response

Body
Name Type Description Required
id String Transaction ID O
model_version String The version of Karlo model that processed the request. O
images Image[] An array of generated images. O
Image
Name Type Description
id String Image ID
seed Integer The seed value used to generate this image.
Used to regenerate an image with the same content.
image String Image file.
A Base64 string or URL depends on the return_type value of the request.
URL is only active for 10 minutes from the response time.
nsfw_content_detected Boolean Whether the image includes NSFW(Not Safe For Work) content.
true: Included
false: Not included

Note: Return null when unused.
nsfw_score Double NSFW score of the image.

Note: Return null when unused.

Sample

Request
Python
curl
# Libraries for REST API requests and image processing
import requests
import json
import urllib
from PIL import Image

# Input the REST API key in [My Application] > [App keys]
REST_API_KEY = '${REST_API_KEY}'

# Generate image
def t2i(prompt, negative_prompt):
    r = requests.post(
        'https://api.kakaobrain.com/v2/inference/karlo/t2i',
        json = {
            'prompt': prompt,
            'negative_prompt': negative_prompt
        },
        headers = {
            'Authorization': f'KakaoAK {REST_API_KEY}',
            'Content-Type': 'application/json'
        }
    )
    # Convert the response to JSON
    response = json.loads(r.content)
    return response

# Prompt
prompt = "A cat with white fur"
negative_prompt = "sleeping cat, dog, human, ugly face, cropped"

# Call the API
response = t2i(prompt, negative_prompt)

# Show the first image of the response
result = Image.open(urllib.request.urlopen(response.get("images")[0].get("image")))
result.show()

curl -v -X POST "https://api.kakaobrain.com/v2/inference/karlo/t2i" \
    -H "Authorization: KakaoAK ${REST_API_KEY}" \
    -H "Content-Type: application/json" \
    -d '{
        "prompt": "A cat with white fur",
        "negative_prompt": "sleeping cat, human, ugly face, cropped"
    }'
Response
HTTP/1.1 200 OK
Content-Type: application/json
{
    "id": "3d6fb820-9845-4b3e-8d6d-7cfd01db5649",
    "model_version": "${MODEL_VERSION}",
    "images": [
        {
            "id": "a80108f8-b9c6-4bf4-aa38-957a38ced4a8",
            "seed": 3878985944,
            "image": "https://mk.kakaocdn.net/dna/karlo/image/..."
        }
    ]
}

Upscale

Basic information
Method URL Authorization
POST https://api.kakaobrain.com/v2/inference/karlo/upscale REST API Key
Permission Prerequisite Kakao Login User consent
- - - -

Upscale the given image.

Include the REST API key in the request header, and send a POST request. Pass the images to upscale in images. Specify the rate to upscale as 2 or 4 by scale.

If failed, refer to Error code to figure out the reason.

Request

Header
Name Description Required
Authorization Authorization: KakaoAK ${REST_API_KEY}
REST API key as a type of user authentication.		 O
Content-Type Content-Type: application/json
The data type of the request.		 O
Body
Name Type Description Required
images String[] An array of the values of image files to upscale, refer to Image file specification.
Images must be with a size of up to 1024 pixels in width and height.
Supports webp, png, jpeg, heic file formats.
(Maximum: 8 images)		 O
scale Integer Rate to upscale the image, 2 or 4.
(Default: 2)

Note: Even if the scale is set to 4, you can only upscale the image up to 2048 pixels in width and height.		 X
image_format String Format to save the image in. One of:
webp
jpeg
png
(Default: webp)		 X
image_quality Integer Save Image Quality.
(Default: 70, Minumum: 1, Maximum: 100)		 X
return_type String Type of the image to return. One of:
base64_string: A Base64 encoded value of the image file
url: A URL of the image file
(Default: url)		 X

Response

Body
Name Type Description
images String[] An array of generated images.
Base64 strings or URLs depends on the return_type value of the request.
URL is only active for 10 minutes from the response time.

Sample

Request
Python
curl
# Libraries for REST API requests and image processing
import requests
import json
import io
import base64
import urllib
from PIL import Image

# Input the REST API key in [My Application] > [App keys]
REST_API_KEY = '${REST_API_KEY}'

# Upscale
def upscale(images, scale=2):
    r = requests.post(
        'https://api.kakaobrain.com/v2/inference/karlo/upscale',
        json = {
            'images': images,
            'scale': scale
        },
        headers = {
            'Authorization': f'KakaoAK {REST_API_KEY}',
            'Content-Type': 'application/json'
        }
    )
    # Convert the response to JSON
    response = json.loads(r.content)
    return response

# Base64 encoding
def imageToString(img):
    img_byte_arr = io.BytesIO()
    img.save(img_byte_arr, format='PNG')
    my_encoded_img = base64.encodebytes(img_byte_arr.getvalue()).decode('ascii')
    return my_encoded_img

# Loading image
img = Image.open('original.png')

# Base64 encoding
img_base64 = imageToString(img)

# An array of Base64 encoded images
img_arr = []
img_arr.append(img_base64)

# Call the API
response = upscale(img_arr, 2)

# Show the first image of the response
result = Image.open(urllib.request.urlopen(response.get("images")[0]))
result.show()

curl -v -X POST "https://api.kakaobrain.com/v2/inference/karlo/upscale" \
    -H "Authorization: KakaoAK ${REST_API_KEY}" \
    -H "Content-Type: application/json" \
    -d '{
        "images": [
            "iVBORw0KGgoAAAANSUhEUgA..."
        ]
    }'
Response
HTTP/1.1 200 OK
Content-Type: application/json
{
    "id": "ca2fff9c-83ae-4def-86b9-efdf107cdb96", 
    "model_version": "${MODEL_VERSION}", 
    "images": [
        "https://mk.kakaocdn.net/dna/karlo/image/..."
    ]
}

Make variation

Basic information
Method URL Authorization
POST https://api.kakaobrain.com/v2/inference/karlo/variations REST API Key
Permission Prerequisite Kakao Login User consent
- - - -

Create variations of the given image.

Include the REST API key in the request header, and send a POST request. Refer to How to use to request more efficiently with optional parameters.

If failed, refer to Error code to figure out the reason.

Request

Header
Name Description Required
Authorization Authorization: KakaoAK ${REST_API_KEY}
REST API key as a type of user authentication.		 O
Content-Type Content-Type: application/json
The data type of the request.		 O
Body
Name Type Description Required
image String A Base64 encoded value of the original image file. O
width Integer Width of the image, a multiple of 8.
(Unit: pixel, Default: 512, Minumum: 384, Maximum: 640)		 X
height Integer Height of the image, a multiple of 8.
(Unit: pixel, Default: 512, Minumum: 384, Maximum: 640)		 X
upscale Boolean Whether to upscale the generated image.
true: Upscale
false: Do not upscale		 X
scale Integer Rate to upscale the image, 2 or 4.
(Default: 2)

Note: Even if the scale is set to 4, you can only upscale the image up to 2048 pixels in width and height.		 X
image_format String Format to save the image in. One of:
webp
jpeg
png
(Default: webp)		 X
image_quality Integer Save Image Quality.
(Default: 70, Minumum: 1, Maximum: 100)		 X
samples Integer Number of images to generate.
(Default: 1, Minumum: 1, Maximum: 8)		 X
return_type String Type of the image to return. One of:
base64_string: A Base64 encoded value of the image file
url: A URL of the image file
(Default: url)		 X
num_inference_steps Integer The number of steps of the decoder denoising process.
(Default: 50, Minumum: 10, Maximum: 100)		 X
guidance_scale Double Guidance scale of denoising process.
(Default: 5.0, Minumum: 1.0, Maximum: 20.0)		 X
scheduler String Scheduler used by decoder denoising process, One of:
decoder_ddim_v_prediction
decoder_ddpm_v_prediction
(Default: decoder_ddim_v_prediction)		 X
seed Integer[] A seed of length equal to the batch size length.
seed must have the same lengh with the number of samples.
Composed of numbers greater than or equal to 0 and less than or equal to 4,294,967,295.
If none, a random.
(Default: null)

Note: Karlo always generates the same image, if all parameters, including seed, have the same value.		 X
nsfw_checker Boolean Whether to perform Detect NSFW content on the generated image.
true: Perform
false Do not perform
(Default: false)		 X
face_refiner FaceRefiner Setting parameters for Refine facial structure. X

Response

Body
Name Type Description
id String Transaction ID
model_version String The version of Karlo model that processed the request.
images Image[] An array of generated images.

Sample

Request
Python
curl
# Libraries for REST API requests and image processing
import requests
import json
import io
import base64
import urllib
from PIL import Image

# Input the REST API key in [My Application] > [App keys]
REST_API_KEY = '${REST_API_KEY}'

# Make variation
def variations(image):
    r = requests.post(
        'https://api.kakaobrain.com/v2/inference/karlo/variations',
        json = {
            'image': image
        },
        headers = {
            'Authorization': f'KakaoAK {REST_API_KEY}',
            'Content-Type': 'application/json'
        }
    )
    # Convert the response to JSON
    response = json.loads(r.content)
    return response

# Base64 encoding
def imageToString(img):
    img_byte_arr = io.BytesIO()
    img.save(img_byte_arr, format='PNG')
    my_encoded_img = base64.encodebytes(img_byte_arr.getvalue()).decode('ascii')
    return my_encoded_img

# Loading image
img = Image.open('original.png')

# Base64 encoding
img_base64 = imageToString(img)

# Call the API
response = variations(img_base64)
print(response)

# Show the first image of the response
result = Image.open(urllib.request.urlopen(response.get("images")[0].get("image")))
result.show()

curl -v -X POST "https://api.kakaobrain.com/v2/inference/karlo/variations" \
    -H "Authorization: KakaoAK ${REST_API_KEY}" \
    -H "Content-Type: application/json" \
    -d '{
        "image": "iVBORw0KGgoAAAANSUhEUgA..."
    }'
Response
HTTP/1.1 200 OK
Content-Type: application/json
{
    "id": "6c0db22f-ecba-4d84-a817-b06e06a207da", 
    "model_version": "${MODEL_VERSION}", 
    "images": [
        {
            "id": "812702f0-3627-4a32-91cc-eaec50910156", 
            "image": "https://mk.kakaocdn.net/dna/karlo/image/...", 
            "seed": 2639508875
        }
    ]
}

Modify image

Basic information
Method URL Authorization
POST https://api.kakaobrain.com/v2/inference/karlo/inpainting REST API key
Permission Prerequisite Kakao Login User consent
- - - -

Creates a connected image based on the given image.

Include the REST API key in the request header, and send a POST request. Refer to How to use to request more efficiently with optional parameters.

If failed, refer to Error code to figure out the reason.

Request

Header
Name Description Required
Authorization Authorization: KakaoAK ${REST_API_KEY}
REST API key as a type of user authentication.		 O
Content-Type Content-Type: application/json
The data type of the request.		 O
Body
Name Type Description Required
prompt String A text describing the desired image.
(Maximum: 2048 characters)		 O
negative_prompt String A text describing anything you want to exclude or control from the image when it is generated.
(Maximum: 2048 characters)		 X
image String A Base64 encoded value of the original image file, refer to Image file specification
Images must be with a size of up to 1024 pixels in width and height.
Supports webp, png, jpeg, heic file formats.		 O
mask String A Base64 encoded value of the image with the area to be edited colored black (RGB or Grayscale value 0), refer to Image file specification O
reference_image String A Base64 encoded value of the image to reference when calibrating the edit area
Images must be with a size of up to 1024 pixels in width and height.
Supports webp, png, jpeg, heic file formats.		 X
upscale Boolean Whether to upscale the generated image.
true: Upscale
false: Do not upscale		 X
scale Integer Rate to upscale the image, 2 or 4.
(Default: 2)

Note: Even if the scale is set to 4, you can only upscale the image up to 2048 pixels in width and height.		 X
image_format String Format to save the image in. One of:
webp
jpeg
png
(Default: webp)		 X
image_quality Integer Save Image Quality.
(Default: 70, Minumum: 1, Maximum: 100)		 X
samples Integer Number of images to generate.
(Default: 1, Minumum: 1, Maximum: 8)		 X
return_type String Type of the image to return. One of:
base64_string: A Base64 encoded value of the image file
url: A URL of the image file
(Default: url)		 X
prior_num_inference_steps Integer The number of steps of the prior denoising process.
(Default: 25, Minumum: 10, Maximum: 100)		 X
prior_guidance_scale Double Guidance scale of the prior denoising process.
(Default: 5.0, Minumum: 1.0, Maximum: 20.0)		 X
num_inference_steps Integer The number of steps of the decoder denoising process.
(Default: 50, Minumum: 10, Maximum: 100)		 X
guidance_scale Double Guidance scale of denoising process.
(Default: 5.0, Minumum: 1.0, Maximum: 20.0)		 X
seed Integer[] A seed of length equal to the batch size length.
seed must have the same lengh with the number of samples.
Composed of numbers greater than or equal to 0 and less than or equal to 4,294,967,295.
If none, a random.
(Default: null)

Note: Karlo always generates the same image, if all parameters, including seed, have the same value.		 X
nsfw_checker Boolean Whether to perform Detect NSFW content on the generated image.
true: Perform
false Do not perform
(Default: false)		 X
face_refiner FaceRefiner Setting parameters for Refine facial structure. X

Response

Body
Name Type Description
id String Transaction ID
model_version String The version of Karlo model that processed the request.
images Image[] An array of generated images.

Sample

Request
Python
curl
# Libraries for REST API requests and image processing
import requests
import json
import io
import base64
import urllib
from PIL import Image

# Input the REST API key in [My Application] > [App keys]
REST_API_KEY = '${REST_API_KEY}'

# Modify image
def inpainting(image, mask):
    r = requests.post(
        'https://api.kakaobrain.com/v2/inference/karlo/inpainting',
        json = {
            'image': image,
            'mask': mask
        },
        headers = {
            'Authorization': f'KakaoAK {REST_API_KEY}',
            'Content-Type': 'application/json'
        }
    )
    # Convert the response to JSON
    response = json.loads(r.content)
    return response

# Base64 encoding
def imageToString(img):
    img_byte_arr = io.BytesIO()
    img.save(img_byte_arr, format='PNG')
    my_encoded_img = base64.encodebytes(img_byte_arr.getvalue()).decode('ascii')
    return my_encoded_img

# Loading image
img = Image.open('image.png')
mask = Image.open('mask.png')

# Base64 encoding
img_base64 = imageToString(img)
mask_base64 = imageToString(mask)

# Call the API
response = inpainting(img_base64,mask_base64)
print(response)

# Show the first image of the response
result = Image.open(urllib.request.urlopen(response.get("images")[0].get("image")))
result.show()

curl -v -X POST "https://api.kakaobrain.com/v2/inference/karlo/variations" \
    -H "Authorization: KakaoAK ${REST_API_KEY}" \
    -H "Content-Type: application/json" \
    -d '{
        "prompt": "tiger",
        "image": "https://mk.kakaocdn.net/dn/karlo-dev/...",
        "mask": "https://mk.kakaocdn.net/dn/karlo-dev/...",
        "negative_prompt": "dog, human",
        "samples": 1,
        "prior_num_inference_steps": 25,
        "prior_guidance_scale": 5,
        "num_inference_steps": 50,
        "guidance_scale": 5,
        "seed": [0],
        "nsfw_checker": true,
        "upscale": false,
        "scale": 2,
        "face_refiner": {
            "bbox_size_threshold": 1,
            "bbox_filter_threshold": 0.5,
            "restoration_repeats": 2,
            "weight_sft": 0.5
        },
        "image_format": "webp",
        "image_quality": 70
    }'
Response
HTTP/1.1 200 OK
Content-Type: application/json
{
    "id": "3d6fb820-9845-4b3e-8d6d-7cfd01db5649",
    "model_version": "${MODEL_VERSION}",
    "images": [
        {
            "id": "a80108f8-b9c6-4bf4-aa38-957a38ced4a8",
            "seed": 3878985944,
            "image": "https://mk.kakaocdn.net/dna/karlo/image/..."
        }
    ]
}

Refine facial structure

Basic information
Method URL Authorization
POST https://api.kakaobrain.com/v2/inference/karlo/face_refiner REST API key
Permission Prerequisite Kakao Login User consent
- - - -

Refines the person's face to be natural in the image.

Note

Generate image, Make variation, Modify image support a parameter to refine facial structure for the result of the request.

Include the REST API key in the request header, and send a POST request. Refer to How to use to request more efficiently with optional parameters.

If failed, refer to Error code to figure out the reason.

Request

Header
Name Description Required
Authorization Authorization: KakaoAK ${REST_API_KEY}
REST API key as a type of user authentication.		 O
Content-Type Content-Type: application/json
The data type of the request.		 O
Body
Name Type Description Required
images String[] An array of the values of image files to upscale, refer to Image file specification.
Images must be with a size of up to 1024 pixels in width and height.
Supports webp, png, jpeg, heic file formats.
(Maximum: 8 images)		 O
image_format String Format to save the image in. One of:
webp
jpeg
png
(Default: webp)		 X
image_quality Integer Save Image Quality.
(Default: 70, Minumum: 1, Maximum: 100)		 X
return_type String Type of the image to return. One of:
base64_string: A Base64 encoded value of the image file
url: A URL of the image file
(Default: url)		 X
bbox_size_threshold Double Ratio of facial area to total image to apply face shape refinement feature, apply feature only if actual facial area is smaller than set size
(Default: 1, Minumum: 0, Maximum: 1)		 X
bbox_filter_threshold Double Threshold to determine if the image is a human face, not applied if 0.
(Default: 1, Minumum: 0, Maximum: 1)		 X
restoration_repeats Double Number of facial structure refinements applied
(Default: 2, Minumum: 1, Maximum: 10)		 X
weight_sft Double Original image retention weight determines how much of the original image is retained.
The higher the value, the more of the original image is preserved.
(Default: 0.5, Minumum: 0, Maximum 1)		 X

Response

Body
Name Type Description Required
id String Transaction ID O
model_version String The version of Karlo model that processed the request. O
images Image[] An array of generated images. O

Sample

Request
Python
curl
# Libraries for REST API requests and image processing
import requests
import json
import io
import base64
import urllib
from PIL import Image

# Input the REST API key in [My Application] > [App keys]
REST_API_KEY = '${REST_API_KEY}'

# Refine facial structure
def faceRefiner(image):
    r = requests.post(
        'https://api.kakaobrain.com/v2/inference/karlo/face_refiner',
        json = {
            'images': [image]
        },
        headers = {
            'Authorization': f'KakaoAK {REST_API_KEY}',
            'Content-Type': 'application/json'
        }
    )
    # Convert the response to JSON
    response = json.loads(r.content)
    return response

# Base64 encoding
def imageToString(img):
    img_byte_arr = io.BytesIO()
    img.save(img_byte_arr, format='PNG')
    my_encoded_img = base64.encodebytes(img_byte_arr.getvalue()).decode('ascii')
    return my_encoded_img

# Loading image
img = Image.open('original.png')

# Base64 encoding
img_base64 = imageToString(img)

# Call the API
response = faceRefiner(img_base64)
print(response)

# Show the first image of the response
result = Image.open(urllib.request.urlopen(response.get("images")[0]))
result.show()

curl -v -X POST "https://api.kakaobrain.com/v2/inference/karlo/face_refiner" \
    -H "Authorization: KakaoAK ${REST_API_KEY}" \
    -H "Content-Type: application/json" \
    -d '{
        "images": [
            "iVBORw0KGgoAAAANSUhEUgA...",
            "aVZCT1J3MEtHZ29BQUFBTNV..."
        ],
        "image_format": "webp",
        "image_quality": 70,
        "bbox_size_threshold": 1,
        "bbox_filter_threshold": 0.5,
        "restoration_repeats": 2,
        "weight_sft": 0.5
    }'
Response
HTTP/1.1 200 OK
Content-Type: application/json
{
    "id": "ca2fff9c-83ae-4def-86b9-efdf107cdb96", 
    "model_version": "${MODEL_VERSION}", 
    "images": [
        "https://mk.kakaocdn.net/dna/karlo/image/...",
        "https://mk.kakaocdn.net/dna/karlo/image/..."
    ]
}

Detect NSFW content

Basic information
Method URL Authorization
POST https://api.kakaobrain.com/v2/inference/karlo/nsfw_checker REST API Key
Permission Prerequisite Kakao Login User consent
- - - -

Detect whether the given image includes NSFW(Not Safe For Work) content.

Note

Generate image, Upscale, Make variation support a parameter to detect NSFW content for the result of the request.

Include the REST API key in the request header, and send a POST request. Pass the images in images.

If failed, refer to Error code to figure out the reason.

Request

Header
Name Description Required
Authorization Authorization: KakaoAK ${REST_API_KEY}
REST API key as a type of user authentication.		 O
Content-Type Content-Type: application/json
The data type of the request.		 O
Body
Name Type Description Required
images String[] An array of Base64 encoded values of the image files.
(Maximum: 8 images)		 O

Response

Body
Name Type Description
id String Transaction ID
model_version String The version of Karlo model that processed the request.
results Result[] The results of NSFW detection
Result
Name Type Description Required
nsfw_content_detected Boolean Whether the image includes NSFW(Not Safe For Work) content.
true: Included
false: Not included		 X
nsfw_score Double NSFW score of the image. X

Sample

Request
Python
curl
# Libraries for REST API requests and image processing
import requests
import json
import io
import base64
from PIL import Image

# Input the REST API key in [My Application] > [App keys]
REST_API_KEY = '${REST_API_KEY}'

# Detect NSFW content
def nsfw(images):
    r = requests.post(
        'https://api.kakaobrain.com/v2/inference/karlo/nsfw_checker',
        json = {
            'images': images
        },
        headers = {
            'Authorization': f'KakaoAK {REST_API_KEY}',
            'Content-Type': 'application/json'
        }
    )
    # Convert the response to JSON
    response = json.loads(r.content)
    return response

# Base64 encoding
def imageToString(img):
    img_byte_arr = io.BytesIO()
    img.save(img_byte_arr, format='PNG')
    my_encoded_img = base64.encodebytes(img_byte_arr.getvalue()).decode('ascii')
    return my_encoded_img

# Loading image
img = Image.open('target.png')

# Base64 encoding
img_base64 = imageToString(img)

# An array of Base64 encoded images
img_arr = []
img_arr.append(img_base64)

# Call the API
response = nsfw(img_arr)

# Print the result
print(response)

curl -v -X POST "https://api.kakaobrain.com/v2/inference/karlo/nsfw_checker" \
    -H "Authorization: KakaoAK ${REST_API_KEY}" \
    -H "Content-Type: application/json" \
    -d '{
        "images": [
            "iVBORw0KGgoAAAANSUhEUgA..."
        ]
    }'
Response
HTTP/1.1 200 OK
Content-Type: application/json
{
    "id": "31aed005-9d31-4248-8a5e-3d086f1eae0e", 
    "model_version": "${MODEL_VERSION}", 
    "results": [
        {
            "nsfw_content_detected": false, 
            "nsfw_score": 0.20849609375
        }
    ]
}
Karlo> REST API