Under the hood it’s just using Tesseract it looks like. Unless you are doing this at large scale I’d look at AWS’s Textract service instead. It does a great job

Under the hood it’s just using Tesseract it looks like. Unless you are doing this at large scale I’d look at AWS’s Textract service instead. It does a great job on PDFs as well as OCR-ing handwriting (cursive in my case). If your data isn’t typed then you can pair this OCR-ing with running the raw text through a LLM to clean it up (I’ve done this and it’s been extremely helpful).