Caveats
- Doesn't work yet with multipass encoding
- Only one master per LAN
- It's usually not a good idea to use MPEG files as source, since the parsing time (which is needed by every client for every segment) can offset the time gained by distributed encoding.
- Distributed encoding is evidently useful only for encoding, so refrain from using it for "Direct Stream Copy"ing video ;-)
- Distributed encoding of the audio stream can be problematic. Thomas Matern only implemented a workaround for mp3 that works by splitting the source audio at silent points to evade audible artifacts. There are three modes from which you can choose in the options menu of the job window.
- If you deactivate "Process Audio", the sound encoding is completely done my the master at the end of the encoding process. "Direct Stream Copy" is not allowed for this mode.
- If you check only "Process Audio", the slaves encode the audio streams themselves, cut at the segement limits choosen by the scheduler.
- If you check both "Process Audio" and "Scan for silence", cutting takes place at the most silent moment. The scanned range depends on several factors. In priciple only the first third of a segment is scanned. To limit memory usage, only 2 minutes (encoded audio) or 1 minute (unencoded audio) is scanned.
This should work independently of the audio source (AVI, MP3, WAV).