Currently Converse doesn't support OCR thereby it can't summarize or Q&A on image based PDFs.

The PDF extraction model needs to be updated to support OCR.