Crop Images for LoRA Training
Prepare image crops for AI datasets. Batch crop images to common AI training sizes such as 512x512, 768x768, and 1024x1024.
AI dataset image cropper
ImageCropKit helps you prepare image crops for LoRA training and AI dataset workflows directly in your browser. Upload multiple images, choose a dataset preset, adjust crops if needed, and download the processed files as a ZIP.
This is useful when your training images need consistent dimensions before being used in a Stable Diffusion or LoRA workflow.
Common LoRA training sizes
Choose from common AI dataset crop sizes:
- 512 × 512 for lightweight SD 1.5 workflows
- 768 × 768 for higher-resolution training sets
- 1024 × 1024 for SDXL-style square datasets
- 1024 × 1536 for portrait images
- 1536 × 1024 for landscape images
Batch crop dataset images
Instead of editing each image one by one, upload your dataset images together and crop them in one batch workflow. You can adjust individual crops and export everything as a ZIP file.
Local processing for dataset privacy
Dataset images are processed locally in your browser. ImageCropKit does not upload your training images to a server, which is useful when working with private or unreleased image sets.
Related image tools
Other ImageCropKit tools that pair well with the one you are using.
Bulk Crop Images
Batch cropping with shared aspect ratio and ZIP download.
Crop Image to 1:1
Make a square image at 512, 768, 1024, 1080, or 2000 pixels.
Crop Image Locally
A private cropper that runs entirely in your browser with no upload.
Crop Image Online
A simple single-image cropper with common aspect ratios.