Home

Awesome

OpenAI has a official API now

Installation

pip install dalle2

Usage

Setup

  1. Go to https://openai.com/dall-e-2/
  2. Create an OpenAI Account
  3. Go to https://labs.openai.com/
  4. Open the Network Tab in Developer Tools
  5. Type a prompt and press "Generate"
  6. Look for fetch to https://labs.openai.com/api/labs/tasks
  7. In the request header look for authorization then get the Bearer Token
from dalle2 import Dalle2
dalle = Dalle2("sess-xxxxxxxxxxxxxxxxxxxxxxxxxxxx") # your bearer key

Generate images

generations = dalle.generate("portal to another dimension, digital art")
print(generations)
✔️ Task created with ID: task-xsuhOthvBXLEjddn3ynyiiOR
⌛ Waiting for task to finish...
...task not completed yet
...task not completed yet
...task not completed yet
...task not completed yet
🙌 Task completed!

[
  {
    'id': 'generation-sCnERSYDPP0Zu14fsdXEcKmL',
    'object': 'generation',
    'created': 1553332711,
    'generation_type': 'ImageGeneration',
    'generation': {
      'image_path': 'https://openailabsprodscus.blob.core.windows.net/private/user-hadpVzldsfs28CwvEZYMUT/generations/generation...'
    },
    'task_id': 'task-nERkiKsdjVCSZ50yD69qewID',
    'prompt_id': 'prompt-2CtaLQsgUbJHHDoJQy9Lul3T',
    'is_public': false
  },
  # 3 more ... 
]

Generate images and download them

file_paths = dalle.generate_and_download("portal to another dimension, digital art")
✔️ Task created with ID: task-xsuhOthvBXLEjddn3ynyiiOR
⌛ Waiting for task to finish...
...task not completed yet
...task not completed yet
...task not completed yet
...task not completed yet
🙌 Task completed!
✔️ Downloaded: C:\...\generation-XySidj4N8EN6Ok9ed15BZ2bs.webp
✔️ Downloaded: C:\...\generation-IK3UdxDz77FA5SLKpQPIITdU.webp
✔️ Downloaded: C:\...\generation-uNejKBXz1z6EQxJAT9pAZbof.webp
✔️ Downloaded: C:\...\generation-Ol1wEqNprf34vNohmJz0iUiE.webp

Generate a specific number of images

generations = dalle.generate_amount("portal to another dimension", 8) # Every generation has batch size 4 -> amount % 4 == 0 works best

# download images
file_paths = dalle.download(generations)
print(file_paths)
✔️ Task created with ID: task-lm0V4nZasgAFasd7AsStE67
⌛ Waiting for task to finish...
...task not completed yet
...task not completed yet
...task not completed yet
...task not completed yet
🙌 Task completed!
✔️ Task created with ID: task-WcetZOHt8asdvHb433gi
⌛ Waiting for task to finish...
...task not completed yet
...task not completed yet
...task not completed yet
...task not completed yet
🙌 Task completed!

Generate images from a masked file

DALL·E supports an "inpainting" API that fills-in transparent parts of an image. The website provides a tool to paint over an existing image to indicate which parts you want to be transparent. This Python package call assumes that the image you provide has already been processed to have transparent parts.

# make the right half of a saved image transparent
from PIL import Image, ImageDraw

image = Image.open('my_image.png')
m, n = image.size

area_to_keep = (0, 0, m//2, n)
image_alpha = Image.new("L", image.size, 0)
draw = ImageDraw.Draw(image_alpha)
draw.rectangle(area_to_keep, fill=255)

image_rgba = image.copy()
image_rgba.putalpha(image_alpha)
image_rgba = image_rgba.resize((1024, 1024)) # image must be square and 1024x1024
image_rgba.save('image_with_transparent_right_half.png')

# ask DALL·E to fill-in the transparent right half
generations = dalle.generate_from_masked_image(
    "portal to another dimension, digital art",
    "image_with_transparent_right_half.png",
)
✔️ Task created with ID: task-xsuhOthvBXLEjddn3ynyiiOR
⌛ Waiting for task to finish...
...task not completed yet
...task not completed yet
...task not completed yet
...task not completed yet
...task not completed yet
🙌 Task completed!

Other languages

Nodejs Package