Prompt_Generator_ZhTW-text-to-image prompt generation.

Example

Input: 'A warrior stands on a cliff, looking out at the sunset over a vast desert.' Output: JSON structure that defines the scene ('warrior on cliff'), environment ('sunset, desert'), and actions ('warrior looking out'). This is then converted into a detailed natural language prompt in English.

Scenario

This function is ideal for users who want to generate highly detailed prompts that a t2i model can interpret clearly. For example, artists working on concept art might describe a scene, and #2pr would refine it into structured English for more nuanced t2i generation.

Example

Input: 'A princess in a grand ballroom, wearing a flowing gown, illuminated by chandeliers.' Output: A list of tags like 'princess, ballroom, chandelier, flowing gown, illuminated', categorized for character, environment, and other key elements.

Scenario

This function is useful for users working with anime or art-style image generation models, where specific tags (like Danbooru tags) are necessary. For example, someone creating anime fan art could use this function to optimize prompts for models that rely on a tag-based system.

Example

Input: 'A lone knight standing in a dark forest.' Output after randomization: 'A wandering bard sitting by a campfire in a dense forest.'

Scenario

Ideal for users who are exploring creative possibilities or want variety in their image generations. This function can introduce randomness while maintaining the core elements of the prompt, allowing users to explore different scenarios or settings, such as game designers brainstorming alternative concepts.