simon is an open source speech recognition suite using the Julius speech recognition engine and the HTK to model the HMM acoustic model.

It does not come with a pre-defined model as the commercial alternatives do, but rather gives the user the possibility to create his own model easily. This essentially means that simon is not bound to a single language, pronunciation or any real limitations. You could teach simon klingon if you wanted to :)

simon can type what it recognizes. However, as creating a dictation model needs LOTS of trainings-data controlling the computer is probably the more realistic use. For that simon provides a plugin-system for commands. Currently it supports the following actions:
Starting Programs
Opening places (folders / urls)
Writing text-macros (snippets)
Controlling the mouse
etc.

simon fully supports Linux and Windows and is available in English and German.