Portrait-aware crop
Detects faces and adjusts their position for a more consistent training dataset.
SwiftCrop AI Dataset Builder
AI Dataset Builder for LoRA, FLUX, SDXL & Pony
Batch crop and resize training images, automatically position portraits, create matching Caption TXT files, and export a consistent dataset without uploading anything to a server.
Free · No account · Local processing · Images + Caption TXT in ZIP
Batch processing
Apply one size, format and quality setting to many images, then export them together as a ZIP.
No image uploads
Detection, cropping, resizing and ZIP creation run locally in your browser.
Caption TXT export
Create one matching text file per image and package the dataset together using ZIP export.
Flexible resolutions
Use 1024×1024, 768×768, 512×512, or any custom size required by your workflow.
Consistent filenames
Generate clean sequential filenames for easier dataset organization and caption pairing.
Choose your dataset workflow
Start with a model-specific guide, then open the same private browser-based SwiftCrop app.
A consistent dataset in three steps
The AI dataset preset selects practical defaults while keeping every size, format, filename and caption setting editable.
Add your images
Use drag and drop, file selection, or paste images directly from the clipboard.
Auto-adjust portraits
SwiftCrop automatically adjusts image composition. You can still fine-tune any image manually.
Export in batch
Create consistent image files, optional matching Caption TXT files, and download the dataset together as a ZIP.
Useful for
- LoRA training image preprocessing
- FLUX, Stable Diffusion and SDXL datasets
- Pony, DreamBooth and portrait datasets
- Bulk square cropping and resizing
Default setup
- Output: 1024 × 1024 px
- JPEG at 95% quality
- Filename: dataset_###
- Optional matching Caption TXT files
- Auto crop after adding images
AI Creator Hub
Explore model-specific dataset preparation guides.
Frequently asked questions
Can I prepare LoRA training images?
Yes. SwiftCrop supports batch cropping, resizing, portrait positioning, and sequential filenames.
Does it work for FLUX, SDXL and Pony?
Yes. It can prepare image datasets for FLUX, Stable Diffusion, SDXL, Pony, DreamBooth, and similar workflows.
Can SwiftCrop export Caption TXT files?
Yes. Enable the AI learning option, enter the caption text, and choose ZIP export to include matching TXT files with the images.
Are my images uploaded or stored?
No. Your images remain on your device and are processed inside the browser.
Do I have to use 1024×1024?
No. You can change the output to 512×512, 768×768, or any custom dimensions in the main tool.
Build a clean AI dataset in minutes.
Open SwiftCrop with the AI dataset preset and finish everything locally in your browser.
Open the free builder