I agree with @geniusquantum :
By far the fastest method is:
-
take your video and split it into chunks
-
copy the chunks to multiple instances
-
do your processing
-
re-assemble the chunks
This can scale to just about any limit, especially if you throw together scripts to automate the process.