LORA Helper V1.1 is a lightweight web app designed to simplify dataset preparation and prompt generation for LoRA training. Powered by JoyCaption, it automatically captions your images, renames them with a consistent prefix, and generates training-ready .txt prompt files — all in one streamlined workflow.
🔑 Key Features
Automated Captioning with JoyCaption
Each image is analyzed and described in natural, detailed language — no manual captioning needed.Checkpoint-Specific Prompt Tailoring
A built-in dropdown menu lets you format descriptions to match the quirks of different models:Default – Clean JoyCaption output
Qwen (Qwen-Image) – Explicit natural language + “Ultra HD/4K” suffix
SDXL – Concise subject, camera/lens, lighting, and editorial style
FLUX – Conversational, plain-sentence descriptions with short negatives
Wan 2.2 – Cinematic film-style phrasing with camera moves
Hunyuan – Structured prompts (subject → details → background → lighting/atmosphere)
Dream (Luma) – Realistic, magazine-style captions with mood and lighting
Customizable Topics
Add optional prompt details (outfit, era, look, body, hair, lighting) to steer the descriptions.Batch Processing
Rename images with a prefix, generate matching.txtfiles, and process entire folders at once.Dark UI with Live Console
Modern interface with a real-time log window so you can monitor progress as images are processed.
🛠️ Why Use It?
Save hours of manual tagging when preparing datasets for LoRA training.
Ensure captions are optimized for the specific checkpoint you’re targeting.
Keep LoRA triggers/keywords properly placed up front for best training results.
Beginner-friendly but flexible enough for advanced users who want fine-tuned prompt styles.
⚡ Whether you’re experimenting with Flux, running high-res SDXL, or styling cinematic datasets for Wan 2.2, LORA Helper adapts your captions automatically so you can focus on creating.