Struggling To Process Loooooooong Document Images With Generative Ai

Upstage added support for very tall document images in its document-parse-250404 release. These images occur in online retail, where a product description may extend for thousands of pixels instead of fitting a conventional page. The company’s announcement describes converting that image into HTML, which a language model can then use to extract named fields and their values. The workflow has two distinct stages: the parser turns the image into a document representation, and a language model extracts information from it. This gives customers a way to turn image-based descriptions into fields that software can use.