Upload
Images and TXT files with matching names. Missing captions can be written in the editor.
Prepare your LoRA dataset
Prepare your images and TXT captions for LoRA training. Edit one caption or tags across a group, crop images and export a ZIP.
Open editor Sign in to upload and save. Uses your existing images.
Review and edit
Filter empty captions, inspect dataset checks, add or replace a tag in selected images. Undo an incorrect bulk edit.
Export
Wait for changes to save, then download images and matching TXT files in a ZIP. Configure training in your chosen trainer.
A clear image + caption pair
mug_01.jpg and mug_01.txt belong together. Keep the subject and visible details accurate. For tags, a bulk replacement can standardize wording while preserving the rest of each caption.
mug_01.jpg + mug_01.txt
Before: ceramic mug, wood table, daylight
Replace: wood table → wooden table
After: ceramic mug, wooden table, daylightArticles and guides
How to caption a LoRA dataset for AI Toolkit
Caption a LoRA dataset for AI Toolkit with character, style and product examples, trigger words, auto-captioning and a review checklist.
How to caption images for LoRA training
Learn how to write LoRA image captions, choose tags or sentences, use a trigger word consistently and edit TXT captions in bulk with Curaset.
A LoRA training dataset example
Explore a LoRA dataset example with images and TXT captions. Match filenames, review captions and check your exported ZIP before training.
Before you upload
Up to 100 files per upload, 20 MiB per file and 500 MiB in total. TXT captions: up to 16 KiB each. Files are stored and processed on the application server under your account.
Curaset edits files; it does not generate images, automatically caption them or train models. Files and storage