Experience gathered while preparing and processing two German co driver sound files has shown that, as Workerbee already said, it's necessary to cut away all silence from individual sound snippets. Don't be shy (I first was) cut like hell and as if you had a razor sharp sword...
We used a free VST plugin (Aradaz Maximizer) for loudness boost which made life a lot easier, if you're not a sound engineer. And remember: downsampling to 11kHz is one of the last sound processing steps.
The final encoding I used for the sound files was 11kHz sample rate, 32bit floating point sample encoding, ogg encoding with max. quality (mostly around 50kbps rate). This matches mostly the original english sound files and doesn't cause any problems or lags or whatever.
If you have a pretty clean recording (i.e. noise level not worse than say -60dB) this is my suggestion for basic sound processing for the plugin:
* use at least 44.1kHz, 32bit for sound processing to avoid clipping and quantization noise while performing maybe a lot of sound processing (depending on your source)
* optional: do some noice level reduction if necessary but pay attention to not introduce much artifacts
* cut out all the sound snippets and name them properly (use named markers and a sound tool that makes your life easier). Audacity should be fine, I used Adobe Audition
* optional: do some sound shaping / folding to kind of intercom sound if you like that; higher pitch voices are to be prefered over low bass voices
* cut off frequencies below 80Hz or more depending on pitch level of recorded voice
* cut off frequencies above 6000Hz, those won't contribute after downsampling to 11kHz in the last step
* normalize all snippets to 0dB or -0.1dB
* boost overall RMS loudness of snippets to somewhere above -5...-3dB RMS(!) or higher if possible (quite some compression needed to achieve this, but otherwise you won't be happy with the loudness of the pacenote calls compared to the other sounds generated by the game)
The above mentioned plugin helped a lot and was very easy to configure for this task. The VST plugin needs stereo input, so converting to stereo before and mono after is a necessary intermediate step. All decent sound tools support batch processing.
* depending on the pitch level of the recorded voice it might be necessary to cut again low and high frequencies after the loudness boost
* downsample to 11kHz, 32bit floating point encoding and convert to ogg file format with some proper tool that can deal with such 32bit encoded source!
Place your generated ogg sound file in a properly named directory under
(RBR path)\Plugins\Pacenote\sounds\(language)\(gender)\(name)
and don't forget to setup your new speaker in the ...\Plugins\Pacenote\PaceNote.ini with a proper entry like
sounds=(your speaker's language)/(gender)/(name)
Fill the parts in () appropriatelly!
Cheers!
Viktor