A command line tool for interacting with DALL-E models, supporting both dall-e-2 and dall-e-3. This project is the little brother of ChatGTP CLI.

Requirements: Python >= 3.10, Pip, Git

Run the following commands:

pip install -U pip
pip install git+https://github.com/marcolardera/dall-e-cli

Then set the OPENAI_API_KEY environment variable with your own OpenAI api key. That's all.

Create an image using dall-e-3 default parameters is easy:

dall-e-cli "A chess player playing on top of Mt. Everest"

Url of the generated image will be printed. Let's try with more custom parameters:

dall-e-cli "A man wearing a banana" --model dall-e-2 --number 3 --size 1024x1024

In this case 3 images of size 1024x1024 will be generated using dall-e-2 and their url printed in the console.

Here is the list of all the available options (also check dall-e-cli --help):

As an alternative to printing the urls of the generated images, DALL-E CLI can download them in a specific folder, using the --download (-d) option:

dall-e-cli "A peaceful town populated by gnomes" -m dall-e-2 -n 4 --download /home/<USERNAME>/ai-images

The output would be somethink like that:

0 -> /home/<USERNAME>/ai-images/1701617525_0.png written
1 -> /home/<USERNAME>/ai-images/1701617525_1.png written
2 -> /home/<USERNAME>/ai-images/1701617525_2.png written
3 -> /home/<USERNAME>/ai-images/1701617525_3.png written

The name of each image is made combining the timestamp with a progressive number.

A command line tool isn't completely useful if it does not support UNIX-style pipelines, right? That's why DALL-E CLI allows to pipe prompts into it using the --pipe (-p) flag. For example:

echo "A WW2 battle taking place on the moon" | dall-e-cli --pipe

For people wondering, here is how it looks like: