Prep Images for LoRA Training

Drop in your photos and get a clean, trainer-ready dataset: JoyCaption analysis on every image, popular-trainer filename patterns, and rich text descriptions tailored to your LoRA goals.

Supports all checkpoints

Scans each photo with JoyCaption (high-quality tags & attributes)

Names files for popular AI training apps (Kohya, AI Toolkit, Civitai Trainer, etc.)

Creates a companion .txt per image with a rich, goal-aligned description

Everything prepped for training

Purpose-built for LoRA creators. Opinionated defaults, trainer-ready outputs.

Checkpoint-agnostic

Compatible with all checkpoints and training flavors. No lock-in—just clean data that slots into your pipeline.

JoyCaption on every image

Per-photo scanning extracts tags, entities, and attributes to guide captions and filename tokens precisely.

Trainer-ready naming

Auto-renames with patterns recognized by popular training apps (Kohya / AI Toolkit / Civitai Trainer and more).

Rich description files

Creates a .txt next to each image with a detailed, goal-aligned description optimized for your LoRA concept.

Constantly updated

Frequent improvements keep up with training best practices and new trainer conventions.

Clean exports

Outputs tidy, de-duplicated folders that are ready to caption, preview, or train immediately.

How it preps your images

Scan with JoyCaption

Each image is analyzed for subject, style, attributes, and context—forming the backbone of accurate captions and filename tokens.

Consistent tag sets per subject
Attribute/pose/style detection
Noise/quality checks

Trainer-ready filenames

Auto-naming uses patterns recognized by leading trainers so buckets, class tokens, and repeats are interpreted correctly.

Kohya / AI Toolkit / Civitai Trainer
Bucket- and repeat-aware schemes
No manual renaming hassles

Rich `.txt` descriptions

Generates a companion text file per image that’s tailored to your LoRA goals—clean, descriptive, and trainer-friendly.

Goal-aligned phrasing
Consistent token ordering
Ready for immediate training

FAQ

Does it support my checkpoint?

Yes. It’s checkpoint-agnostic and works across all popular LoRA training setups.

How are captions created?

Each photo is scanned with JoyCaption to extract accurate tags and attributes, then written into a rich .txt tailored to your LoRA goals.

Which trainer naming schemes are supported?

The tool names files for all popular training applications, including Kohya, AI Toolkit, and Civitai Trainer—so buckets and repeats are read correctly.