DrawThings Image CLI

The DrawThings Image CLI (dtcli) is a command-line tool that is used to generate images on devices running the Mac and IOS app DrawThings.

This tool is built using Python and the Click library, making it easy to use and configure. This was my first real Python project so there may be some lumps 😆

Installation

To install the DrawThings Image CLI you have 2 options, follow these steps:

Option 1

1. Clone this repository to your local machine.

2. Install poetry (refer to their documentation)

3. Navigate to the project directory and open the terminal at said location:

4. Install the required dependencies using the followping:

   poetry install
   

Option 2

1. Install pipx via their simple documentation

2. Run pipx install git+https://github.com/O-J1/drawthings-image-cli.git

3. If everything goes well you will see:

   These apps are now globally available — dtcli.exe done! ✨ 🌟 ✨

This means that dtcli is now a proper command line tool that you can use independently of Poetry and is isnt polluting your global python install thanks to pipx!

Usage

The DrawThings Image CLI provides just one command and many optional "options" for generating images. Here's how you can use it:

Basic Command

To generate an image using default settings, run the following command:

poetry run dtcli

P.S Please note that this project will output into a folder named "DrawThings Output" that is under the current working directory that the CLI command is executed at.

The order of priority for options are as follows:

CLI inputs > dt_config.toml > default values from constants.py

Options

You can customize the image generation process using various options:

• -u, --url: Specify the target device URL for image generation.
• -p, --prompt: Set the prompt for the image generation.
• -n, --negative_prompt: Specify the negative prompt for the image generation.
• -m,--model: Specify what diffusion model to use, you may use the file name or the display name seen in the app

For example:

poetry run dtcli --url "192.168.1.100:7860" --prompt "landscape, fantasy" -n "low quality, worst quality, monochrome"

Poetry run can be excluded if you went with option 2 for install.

Advanced Commands

The DrawThings Image CLI provides more advanced commands:

• -c,--config: If you choose to not use the .toml created in the current working directory you may specify a path
• -s, --steps: Adjust the steps, however it already uses the reccomended value of 30, in the large majority of cases there should be no need to adjust
• --seed: Set the seed for noise generation, defaults to random
• --sampler: Adjust the sampler. Consider using only Euler a, DPM++ 2M Karras and DPM++ SDE Karras with the latter needing only 20 steps. Defaults to Euler A
• -bsize, --batch_size: Adjusts the concurrent image generation (max of 4)
• -bcount, --batch_count: Adjusts the consecutive image generation, think of it like a loop (max of 100)

Configuration

You can configure default options by edit the dt_config.toml file created in the current working directory or manually creating it somewhere on your PC and passing the path via -c or --config. Here's an example of the contents:

url = "http://192.168.1.100:7860"
timeout = 120
prompt = "(limited palette blue black:1.1), creative (command line interface CLI:1.3) connected to a (blue magical tesseract glowing cube:1.2), no humans"
negative_prompt = "human, man, woman, child, simple background, (low quality, worst quality, normal quality:1.4), hand"
seed = -1
steps = 30
sampler = "Euler a"
batch_size = 1
batch_count = 1
guidance_scale = 7
hires_fix = false
default_model_list = [
    "Generic (Stable Diffusion v1.5)",
    "SDXL Base (v1.0)",
]
manual_model_list = []

You can read more about TOML syntax here: https://toml.io/en/

Its a good middle ground between INI and JSON.

Contributing

Contributions to this project are welcome! If you find issues or have improvements, feel free to create a pull request.

License

This project is licensed under AGPL_V3.

Acknowledgments

This project interfaces with DrawThings for remote image generation. We are just a shell/client for this great tool.

Special thanks to the following individuals:

• Inspiriation from Bob K's edit of the A111 API script
• Scyren for assistance for providing me with some good hints on resolving some of moderate callback issues
• Huge thanks to L3viathan who without their help I couldnt have easily patched the arg parser for click
• Tachyondecay for pointing me in the right direction for click documentation
• Liu Liu for the work on the DrawThings app. Read more about some of the work he has done @ https://engineering.drawthings.ai/
• Clarityai engeering for their guide on using poetry + pipx to create a CLI tool @ https://medium.com/clarityai-engineering/how-to-create-and-distribute-a-minimalist-cli-tool-with-python-poetry-click-and-pipx-c0580af4c026