hi, just a short description about this: the rtpproxy-transcoder implements another command (I think it was 't') which tells rtpproxy to transcode the stream. The codec must be selected by the force_transcode function in nathelper and passed to the rtpproxy-transcoder. With the current implementation the selected codec is fixed. This could be improved of course to make it choosing a codec from priority list or the like. The rtpproxy-transcoder uses codec modules from SEMS. Any codec which is implemented in SEMS can be used by the transcoder. So if there is a g927 implementation for SEMS (not yet) you could use it. The transcoder just invokes the encode and decode functions of the codec module. For adaptation of different packet sizes it also buffers audio until there is enough audio to be sent. It also detects whether there is packet loss and in that case throws away old buffered audio. If this is added to rtpproxy as it is (without #defines), every normal not transcoded rtpproxied session has two unused empty pointers. maybe we could live with that. Stefan Atle Samuelsen wrote: -- Stefan Sayer Media Services Development stefan.sayer(a)iptego.de www.iptego.de iptego GmbH Am Borsigturm 40 13507 Berlin Germany Amtsgericht Charlottenburg, HRB 101010 Geschaeftsfuehrer: Alexander Hoffmann