When querying the API for audio thumbnails (yeah, appears to be stupid ...) it returns a height and width in a manner that let the thumb look distorted.
Result:
"thumburl": "https://bits.wikimedia.org/static-1.25wmf22/resources/assets/file-type-icons/fileicon-ogg.png",
"thumbwidth": 120,
"thumbheight": 13,
https://bits.wikimedia.org/static-1.25wmf22/resources/assets/file-type-icons/fileicon-ogg.png is a 120x120px file which is then squashed to 13px height