Standalone util to caption images for lora or stable-diffusion training. No server required - fully local, modifiable, minimalistic.
Flow:
- Select images
- Modify existing embedded prompt (only for png with tEXt Textual embedding - e.g. Automatic1111 generated images)
- Iterate over all images
- Export the captions (naming of .txt always respective to img file)
- Use captions and images with kohya or any other training framework
Demo
Exported captions
$ npm install$ npm run devopen localhost:3000
Images used for demo