• The engine and the developement toolkit, supports very large vocabulary and morphlogically rich languages using statistical language models. 

• Acoustic model is speaker independent but can be adapted to the speaker or to special acoustic conditions. 

• Thematic transcription for continuous speech recognition, which is adaptable to several topics like call center analytics, parliamentary speech recognition, broadcast news indexing, etc. 

• Low latency real-time engine facilitates online closed captioning. 

• Possible applications include indexing, categorization and analysis of large media archives.