This tool leverages the DallE model via the OpenAI API to generate images in batches. It's designed for users looking to create image datasets for purposes such as concept fine-tuning or style training for AI image models, including Stable Diffusion.
Follow these steps to use the DallE-Image-Generator:
- Navigate to the
settings.inifile and input your OpenAI API Key. - Run
DallEGenerator.py. - Familiarize yourself with the settings on the top/right side of the interface. Hover over each setting for more information, or refer to the detailed settings section below.
- Enter your image generation prompt. For dynamic prompts, you can incorporate [VARIABLES] within brackets.
- (Optional) If using [VARIABLES], press the
Analyze Promptbutton. This will introduce a text field for inputting values for each variable. This step is necessary only if your prompt includes variables. - (Optional) For each [VARIABLE], provide a list of variable values. Each entry generates one image. Note: Utilizing multiple variables results in the creation of all possible combinations, which can significantly increase the number of generated images. Exercise caution with this feature.
- Click
Preview Promptto review your prompt configuration. - Hit the
Generatebutton. You will receive an estimate of the cost involved and will be asked to confirm before proceeding. - The generated images, along with captions and logs if selected, will be saved in a folder named after today's date. Organization within this folder depends on your chosen settings.
Important Notice: After modifying settings, you must save your changes using the [SAVE SETTINGS] button and restart the program before generating new images. Due to unresolved interface bugs, a program restart is necessary to ensure settings are applied correctly.
Dataset: Designates a sub-folder name within the output folder where all images will be stored. If combined with the Caption option, this name will prefix the caption file.
Caption: Enables the output of a caption file. If a dataset name is provided, it prefixes the caption. The caption content is derived from the provided prompt or, if [conceptify] is enabled, from the concept name.
Log: Generates a .log file for each batch of images, containing the prompt and other relevant information.
Conceptify: When activated, searches for an additional set of brackets in [VARIABLES] to use as a concise descriptor for organizing images into sub-folders and generating captions accordingly.
Quantity: Specifies how many copies of each image to generate. It is recommended not to exceed 9 to avoid excessive generation.
Model Version: Select between DallE2 and DallE3 models.
Quality: Choose between Standard and HD quality, the latter being available only for DallE3.
Resolution: Determines the resolution of the generated images, contingent on the chosen model.
The example below illustrates generating 12 images inspired by Batman's style, encompassing variations with an airplane, a bookshelf, and a coffee machine. These images will be organized in a 'BatmanStyle' main folder, with respective sub-folders for each category. A caption file will accompany each category, prefixed with 'BatmanStyle' and the specific category as the caption.
