IMG Dataset Refiner v4.3 Pro auto-captions and preps LoRA datasets

IMG Dataset Refiner v4.3 Pro transforms dataset preparation for AI model training with a major update. Originally a visual manager and balancer, the tool now provides full AI integration for auto-captioning, translation, and hallucination detection via local engines (LM Studio, Ollama) or cloud APIs (Claude, Gemini, OpenAI). It also introduces a Smart AI Recipe Generator that analyzes your entire dataset and produces an optimized keyword recipe—pinning the trigger word at the top—for easy upload to Civitai. A mass batch editor lets you add, remove, or replace tags across thousands of images in one click.

The tool includes built-in preprocessing features: a visual duplicate finder, smart face cropping, and high-quality bulk resizing. The UI has been redesigned for speed with native drag-and-drop for Windows folders, side toggles for a larger workspace, and real-time translation support. It remains 100% open-source and now ships with 1-click Windows install scripts, eliminating the need to touch the terminal. Ideal for Flux, SD3, and SDXL LoRA training, this update turns the project into a complete data engineering suite for AI practitioners.