Do you have long document images that you want to process with generative AI but can't find the right solution? You're not alone. Many industries rely on vertically long images, particularly in retail and e-commerce across Korea, Japan, and China, where product descriptions often span thousands of pixels in length.

However, when our customers benchmarked various document processing solutions, they encountered two major issues:

1. Many products simply do not support long images.
2. Those that do often suffer from a drastic drop in quality compared to standard-sized images.

You asked, and we delivered.

Long image parsing with Upstage Document Parse

‍

With document-parse-250404, Upstage Document Parse now supports extremely long image processing—achieving a 38.596% improvement in accuracy over the previous version.

How our customers are using this feature

With the upgraded Document Parse, our customers go beyond simple document parsing by leveraging their choice of LLM (ideally Solar 🙂). Since the resulting HTML is highly accurate, they can extract high-quality key-values with ease. Here’s an example of a common workflow:

At Upstage, we know that the document universe is vast, with countless edge cases that existing solutions struggle to handle. Do you have a pain point when processing your documents? We’re here to listen.

Contact us to share your challenges, or try the latest Document Parse yourself in our playground.

‍

‍