Hi Streamlit community! 
I built a small Streamlit app as a proof-of-concept for a Korean document OCR workflow.
It started as a bankbook document validation demo: select sample documents, run OCR, and compare the recognized text with expected vendor and bank information.
I recently added a simple “Try with your own image” feature, so the public demo can also be used with any PNG/JPG containing text.
Upload → Resize → Deskew → PaddleOCR → Text
The deskew preprocessing is also separated as a reusable Python utility and reused by the interactive demo.
The UI and sample documents are in Korean, but the source code and processing flow should be fairly easy to follow.
Live Demo: https://vibe-coding-bankbook-dummy.streamlit.app/
Source: bankbook_dummy/README.md at main · nevlock59-langlock/bankbook_dummy · GitHub
I’d especially appreciate feedback on the Streamlit UI/UX and how the demo feels to use!
