I think you need to provide more detail on the limits - for example, does the output in converted.txt need to be in the same order as 'before.txt', how long does each parse of the python script take ? If the output order is not contingent on the input, you may be able to do this by backgrounding the processes and launching a number of them in each loop - the number depending, I guess, on how many threads your CPU will handle.
Something like the following might (or might not) suit your purpose:
#! /bin/bash threads=4; threads=$(( $threads - 1)) while read filein do python py-hex.py $filein | sed -n -e '/^qter/p' | sed 's/qter: //g' >> converted.txt & for thread in `seq $threads` do read filein python py-hex.py $filein | sed -n -e '/^qter/p' | sed 's/qter: //g' >> converted.txt & done done < $filename Notes: This assumes your python file can handle empty inputs (ie if the number of commands not exactly divisible by the number of threads there will be some empty lines - you could always do a check for this before executing the inner loop.
This script assumes you don't care about the output order.