This workflow is designed to intelligently generate and refine prompts for SDXL using a combination of tag-based input, large language model (LLM) processing, and structured transformations. It extracts and converts character and scene details from image metadata or user-provided tags, transforming them into natural-language descriptions. The workflow then enhances the prompt by incorporating dynamic scene elements, adjusting descriptions for clarity, and ensuring structured output. Additionally, it offers a reverse conversion feature to turn refined prompts back into structured SDXL-friendly tags. Ideal for automating and improving the quality of AI-generated pony-themed image prompts.

LLM Model: https://huggingface.co/bartowski/NemoMix-Unleashed-12B-GGUF/blob/main/NemoMix-Unleashed-12B-IQ4_XS.gguf?language=python