I got sick of the existing excerpt functions not handling posts without spaces (i.e. in Japanese / Chinese / Korean). This plugin will excerpt using the following priority:
* try to break on a sentence boundary
* try to break on a space
* break on a fixed character length
It is fully UTF-8 aware and will not break UTF-8 characters (note, it assumes that the blog uses UTF-8 encoding).
Bare details and download at:
http://code.jellycan.com/ml-fancy-excerpt/