Renders large text blobs as dense PNG image pages via pxpipe, enabling vision models to read content at ~3x token efficiency with a single inline image block.
