With the YouTube clamp down on copyright and copyleft, it would be a doable and quick hack to integrate what I developed for a site in 2002 - a 'visual tag cloud'. For this good ol' perl or now python is appropriate and does the job with a tad of help from imagemagick.

A visual tag cloud would provide thumbs of CC pictures from Wikimedia etc. that are tagged with content that matches a CJists story content. This would make it a lot quicker for CJists to integrate visuals into their Highlighted or self-written stories.

Eventually there would be an CC imagebank at NowPublic for future stories. This could even be done now with an agregator working on tags. The major drawback is lazy folksonomy and mistaged images. Wikipedia is probably the best source for now but many public and CC images from photo-sharing sties could be scaped and indexed and with the use of image recognition softare, faces and areas could be tagged by NowPublic Cjists building over time...

I can think of other ways to improve and ascertain that copyright problems do not become an issue at NP and that still the stories are well illustrated. There will be need for sure of web2.0 photo-editing, never is there a job loss from intelligent web2 services. Student interns could cut their teeth on such a task, and of course, pictures are fun.