You need a de-esser. What are you using (for a recording medium)? Maybe there's a plug you can use...
Otherwise, a sidechained compressor (which is what a de-esser plugin is, generally). You mult the vocal track through a graphic EQ, pull everything down except for the sibilance (generally, you'll have a hump in the graph from about 4-8K or so). Plug the output of the graphic into the sidechain input on the compressor (and the regular vocal track through the regular in/out).
The sibilance will then trigger the compressor - You'll need to adjust your settings to get it to compress only the sibilance. You're attack and release will need to be fairly quick.