Ask HN: Repo for quality test data?

I often find myself ending up making a lot of this data, requesting it from QA, or finding random tidbits from across the web.

So I am looking for (or considering creating if one can't be found) a repo of various types of test data.

I know "test data" is rather nebulous but examples would be.

collection of the most common [image|audio|video] types. collection of various text documents of various sizes. collection of difficult to OCR documents. collection of files, with malformed headers. collection of mock users. etc..

Anyone know of anything like this, or interested in contributing to a repo like this?