Mode: SD-Lora Assistant
# General-Purpose LoRA Dataset Preparation Workflow This is the core prompt for the **General-Purpose LoRA Dataset Preparation Workflow**, an automated system designed to transform any of your training concepts into a high-quality, precisely tagged dataset. Whether it's "drones," "sports cars," or "architectural styles," you just need to provide a simple configuration, and the system will automatically handle the tedious work of image retrieval, analysis, and tagging. Your dataset will no longer be a chaotic collection of pictures, but rather a precisely empowered, valuable resource ready for training. **Core Features:** * **Dynamic Requirement Definition:** Completely get rid of hard-coding. Define **any theme** through a flexible `user_config` object, including its **base tags**, **sub-types**, and **related scenarios**. * **Cross-Platform Smart Retrieval:** The system automatically generates dozens of precise search queries based on your configuration and executes searches on both Google and Bing simultaneously to ensure image **diversity** and **high quality**. * **AI Deep Image Analysis:** For every retrieved image, the workflow invokes an AI vision model to conduct a **deep analysis** based on your defined theme, accurately describing the image's **attributes**, **actions**, **environment**, and **perspective**. * **Standardized Tag Generation:** Automatically converts the AI's natural language analysis results into the clean, consistent, **comma-separated tags** required for LoRA training (e.g., `side view, flying, blue sky, over water`). * **Efficient Batch Processing & Caching:** Automatically breaks down large tasks (e.g., 1000 images) into manageable small **batches** (e.g., 5 at a time) for processing, and **caches** retrieval results to ensure the process is stable, efficient, and can be interrupted/resumed at any time.
Mode: SD-Lora Assistant
# General-Purpose LoRA Dataset Preparation Workflow This is the core prompt for the **General-Purpose LoRA Dataset Preparation Workflow**, an automated system designed to transform any of your training concepts into a high-quality, precisely tagged dataset. Whether it's "drones," "sports cars," or "architectural styles," you just need to provide a simple configuration, and the system will automatically handle the tedious work of image retrieval, analysis, and tagging. Your dataset will no longer be a chaotic collection of pictures, but rather a precisely empowered, valuable resource ready for training. **Core Features:** * **Dynamic Requirement Definition:** Completely get rid of hard-coding. Define **any theme** through a flexible `user_config` object, including its **base tags**, **sub-types**, and **related scenarios**. * **Cross-Platform Smart Retrieval:** The system automatically generates dozens of precise search queries based on your configuration and executes searches on both Google and Bing simultaneously to ensure image **diversity** and **high quality**. * **AI Deep Image Analysis:** For every retrieved image, the workflow invokes an AI vision model to conduct a **deep analysis** based on your defined theme, accurately describing the image's **attributes**, **actions**, **environment**, and **perspective**. * **Standardized Tag Generation:** Automatically converts the AI's natural language analysis results into the clean, consistent, **comma-separated tags** required for LoRA training (e.g., `side view, flying, blue sky, over water`). * **Efficient Batch Processing & Caching:** Automatically breaks down large tasks (e.g., 1000 images) into manageable small **batches** (e.g., 5 at a time) for processing, and **caches** retrieval results to ensure the process is stable, efficient, and can be interrupted/resumed at any time.
How To Use
