Learn the basics of the Recraft API, including raster and vector image generation, style creation, image generation in your brand style and colors, image vectorization and background removal.
Authenticate and interact with our API in a matter of minutes.
Understand our flexible API pricing plans.
Dig into the details of using the Recraft API.
Practical guides for both CLI and Python usage.
A style is a descriptor that defines the visual appearance and feel of generated images. This includes a wide array of elements such as textures and visual effects, shapes and colors, composition and lines, etc. There are four classes of base styles: realistic image, digital illustration, vector illustration and icon.
Images of style “realistic” are expected to look like just ordinary photographs made with a digital camera or a smartphone or a film camera.
Images of style “digital illustration” are pictures drawn by hand or using computers - virtually everything except photos and vector illustrations. The most crucial difference from “realistic images” is that illustrations possess simplified textures (like in 3D-rendered or manually drawn images) - or they are stylized in a certain creative way. The difference from “vector illustration” is that “digital illustrations” allow for more complex color transitions, shades, fine textures.
Images of style “vector illustration” are expected to look like those drawn using vector graphics (see Wikipedia). Usually, they use only a few different colors at once, shapes are filled with flat colors or simple color gradients. Shapes of objects can be arbitrarily complex.
Images of style “icon” are small digital images or symbols used in the graphical user interface. They are designed to be simple and recognizable at small sizes, often visually summarizing the action or object they stand for, or they can act as the visual identity for an app or a website and are crucial in branding.
A style can be refined by adding a substyle for more precise definition. Below, you will find a list of supported styles and corresponding substyles. Please note that the available styles and substyles may vary depending on the model. Additionally, you have an option to create your own style by combining a base style (e.g. realistic image, digital illustration, vector illustration or icon) with a collection of reference images.
Recraft developed two powerful models: Recraft20B and RecraftV3.
The Recraft20B model was released in February 2024 and was the first AI model built specifically for designers. It allowed them to create both vector and raster images that are anatomically perfect, maintain consistent brand style and also iterate with ultimate control and precision.
In October 2024 the new model was introduced - Recraft V3. Our team trained a new SOTA model from scratch and set a new standard for excellence in image generation. Recraft V3 participated in the Hugging Face’s industry-leading Text-to-Image Model Leaderboard by Artificial Analysis. It secured #1 place with ELO rating of 1172.
Now both models are available through the Recraft API.
The following are the API Unit packages available from Recraft for the use of the Recraft API Service. API Unit packages must be purchased in advance and all API Unit packages are non-cancellable and non-refundable. Any number of unit packages can be bought.
The following are the Service charges (in API Units) for use of the API Services. API Units will be automatically deducted from Member’s pre-purchased API Unit package upon use of the described Service by the Member. The Recraft Service is the system of record for determining API Unit usage. Please note, Services charges do not depend on model.
Authenticate and interact with our API in a matter of minutes.
We use Bearer API tokens for authentication. To access your API key, log in to Recraft, enter your profile and hit 'Generate' (available only if your API units balance is above zero). All requests should include your API key in an Authorization HTTP header as follows:
Authorization: Bearer RECRAFT_API_TOKEN
The Recraft API adheres to REST principles, allowing you to interact using any utilities (e.g., curl), programming languages, or libraries of your choice.
One of the easiest of available alternatives is OpenAI Python library which is also compatible with Recraft API, but it’s important to remember that not all parameters/options are supported or implemented. Additionally, some parameters may have different meanings, or they may be quietly ignored if they are not applicable to the Recraft API.
Future examples will be shown using that library, for example, once installed, you can use the following code to be authenticated:
from openai import OpenAI
client = OpenAI(
base_url='https://external.api.recraft.ai/v1',
api_key=<TOKEN>,
)
Dig into the details of using the Recraft API.
Creates an image given a prompt.
POST https://external.api.recraft.ai/v1/images/generations
response = client.images.generate(
prompt='race car on a track',
style='digital_illustration',
)
print(response.data[0].url)
URL: https://img.recraft.ai/TH4CG_8z2Ql8G1B_mx9tkVQk67i6TmdM9Av13L2-ERE
/rs:fit:1024:1024:0/raw:1/plain/abs:/
/prod/images/a9f881cb-7638-4fa8-88d2-cc65dbfd0375
Note: style_id
and style
parameters are mutually exclusive. If neither of the two parameters is specified, the default style of realistic_image
will be used
Hint: if OpenAI Python Library is used, non-standard parameters can be passed using the extra_body
argument. For example:
response = client.images.generate(
prompt='race car on a track',
extra_body={
'style_id': style_id,
'controls': {
...
}
}
)
print(response.data[0].url)
Converts a given raster image to SVG format.
POST https://external.api.recraft.ai/v1/images/vectorize
response = client.post(
path='/images/vectorize',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
files={'file': open('image.png', 'rb')},
)
print(response['image']['url'])
https://img.recraft.ai/fZm6nwEjI9Qy94LukIKbxRm4w2i5crwqu459qKg7ZWY
/rs:fit:1341:1341:0/raw:1/plain/abs://external/images/2835e19f-282b-419b-b80c-9231a3d51517
Body of a request should be a file in PNG format and parameters passed as content type 'multipart/form-data'
.
Removes background of a given raster image.
POST https://external.api.recraft.ai/v1/images/removeBackground
response = client.post(
path='/images/removeBackground',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
files={'file': open('image.png', 'rb')},
)
print(response['image']['url'])
https://img.recraft.ai/EYOLjpky-2-uClelfP61kzK-SEpIhKgLfjLFFGxmM_U
/rs:fit:0:0:0/raw:1/plain/abs://external/images/e2d0cba6-12df-4141-aa21-43bfd5889990
Body of a request should be a file in PNG format and parameters passed as content type 'multipart/form-data'
.
Enhances a given raster image using ‘clarity upscale’ tool, increasing image resolution, making the image sharper and cleaner.
POST https://external.api.recraft.ai/v1/images/clarityUpscale
response = client.post(
path='/images/clarityUpscale',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
files={'file': open('image.png', 'rb')},
)
print(response['image']['url'])
https://img.recraft.ai/LtCo_bs3chC8zhrku0CWLpCBKv4iOODprEdeD_MY1dw
/rs:fit:1760:2348:0/raw:1/plain/abs://external/images/f7d01b15-0eba-4439-a5fb-38af38fb524e
Body of a request should be a file in PNG format and parameters passed as content type multipart/form-data
.
Enhances a given raster image using ‘generative upscale’ tool, boosting resolution with a focus on refining small details and faces.
POST https://external.api.recraft.ai/v1/images/generativeUpscale
response = client.post(
path='/images/generativeUpscale',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
files={'file': open('image.png', 'rb')},
)
print(response['image']['url'])
https://img.recraft.ai/DV4d9pMeq5lIluqS7m8qHyg-mb6hf5uCqEPPC8t8wy4
/rs:fit:4740:3536:0/raw:1/plain/abs://external/images/fb576169-8a66-4270-a566-35713ad72020
Body of a request should be a file in PNG format and parameters passed as content type multipart/form-data
.
Upload a set of images to create a style reference.
POST https://external.api.recraft.ai/v1/styles
response = client.post(
path='/styles',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
body={'style': 'digital_illustration'},
files={'file1': open('image.png', 'rb')},
)
print(response['id'])
{"id": "229b2a75-05e4-4580-85f9-b47ee521a00d"}
Upload a set of images to create a style reference.
The generation process can be adjusted with a number of tweaks.
Color type is defined as an object with the following fields
response = client.images.generate(
prompt='race car on a track',
style='realistic_image',
extra_body={
'controls': {
'image_type': 'realistic_image',
'colors': [
{'rgb': [0, 255, 0]}
]
}
}
)
print(response.data[0].url)
Generate AI images using cURL or Python and create your own styles programmatically.
curl https://external.api.recraft.ai/v1/images/generations \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-d '{
"prompt": "two race cars on a track",
"style": "digital_illustration",
"model": "recraftv3"
}'
from openai import OpenAI
client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN)
response = client.images.generate(
prompt='two race cars on a track',
style='digital_illustration',
model='recraftv3',
)
print(response.data[0].url)
curl https://external.api.recraft.ai/v1/images/generations \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-d '{
"prompt": "red point siamese cat",
"style": "realistic_image",
"size": "1280x1024"
}'
from openai import OpenAI
client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN)
response = client.images.generate(
prompt='red point siamese cat',
style='realistic_image',
size='1280x1024',
)
print(response.data[0].url)
curl https://external.api.recraft.ai/v1/images/generations \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-d '{
"prompt": "a monster with lots of hands",
"style": "digital_illustration",
"substyle": "hand_drawn"
}'
from openai import OpenAI
client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN)
response = client.images.generate(
prompt='a monster with lots of hands',
style='digital_illustration',
extra_body={'substyle': 'hand_drawn'},
)
print(response.data[0].url)
curl -X POST https://external.api.recraft.ai/v1/styles \
-H "Content-Type: multipart/form-data" \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-F "style=digital_illustration" \
-F "file=@image.png"
# response: {"id":"095b9f9d-f06f-4b4e-9bb2-d4f823203427"}
curl https://external.api.recraft.ai/v1/images/generations \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-d '{
"prompt": "wood potato masher",
"style_id": "095b9f9d-f06f-4b4e-9bb2-d4f823203427"
}'
from openai import OpenAI
client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN)
style = client.post(
path='/styles',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
body={'style': 'digital_illustration'},
files={'file': open('image.png', 'rb')},
)
print(style['id'])
response = client.images.generate(
prompt='wood potato masher',
extra_body={'style_id': style['id']},
)
print(response.data[0].url)
curl -X POST https://external.api.recraft.ai/v1/images/vectorize \
-H "Content-Type: multipart/form-data" \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-F "file=@image.png"
from openai import OpenAI
client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN)
response = client.post(
path='/images/vectorize',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
files={'file': open('image.png', 'rb')},
)
print(response['image']['url'])
curl -X POST https://external.api.recraft.ai/v1/images/removeBackground \
-H "Content-Type: multipart/form-data" \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-F "response_format=b64_json" \
-F "file=@image.png"
from openai import OpenAI
client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN)
style = client.post(
path='/images/removeBackground',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
body={'response_format': 'b64_json'},
files={'file': open('image.png', 'rb')},
)
print(response['image']['url'])
curl -X POST https://external.api.recraft.ai/v1/images/clarityUpscale \
-H "Content-Type: multipart/form-data" \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-F "response_format=b64_json" \
-F "file=@image.png"
from openai import OpenAI
client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN)
style = client.post(
path='/images/clarityUpscale',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
body={'response_format': 'b64_json'},
files={'file': open('image.png', 'rb')},
)
print(response['image']['url'])
curl -X POST https://external.api.recraft.ai/v1/images/generativeUpscale \
-H "Content-Type: multipart/form-data" \
-H "Authorization: Bearer $RECRAFT_API_TOKEN" \
-F "file=@image.png"
from openai import OpenAI
client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN)
style = client.post(
path='/images/generativeUpscale',
cast_to=object,
options={'headers': {'Content-Type': 'multipart/form-data'}},
files={'file': open('image.png', 'rb')},
)
print(response['image']['url'])
- 1024x1024
- 1365x1024
- 1024x1365
- 1536x1024
- 1024x1536
- 1820x1024
- 1024x1820
- 1024x2048
- 2048x1024
- 1434x1024
- 1024x1434
- 1024x1280
- 1280x1024
- 1024x1707
- 1707x1024