Earlier this month we discussed how the VGKG has historically used highly optimized image libraries that allowed us to absolutely minimize the computational resources required to process each image, but which have increasingly struggled with evolving image standards and increasing use of previously rarer format extensions. Over the coming 48 hours we will be transitioning the VGKG to the new GEN4 architecture.

As part of this transition, images that fail all other attempts will now be converted through ImageMagick. While ImageMagick is an extremely heavyweight application, it offers unrivaled support for the full range of features of each image format, including less common extensions. It is also typically reasonably robust to lightly damaged images and best replicates the "best effort" loading of modern web browsers.

We are also relaxing the maximum size of images that we process to handle the increasing use of unoptimized high-resolution imagery by news outlets that exceeds the maximum dimension limits of the GCP Cloud Vision API. In those cases we adaptively downsize the image to the requirements of the API.

You will see this reflected in the VGKG's makeup moving forward, as higher resolution and more boundary-pushing image format uses are included.