Actually, there is a neater way to visualize Natural Bond Orbitals than I asked for in feature request 2974999.
When PLOT (it has no subparameters) is specified in the NBO parameter list, all files needed for visualization are produced. The NBO parameter list ($NBO ... $END) appears in the electronic structure system input if NBO is embedded in the ESS, and appears in the .47 input file to GenNBO if NBO is being run standalone. $NBO AONBO=P $END is not necessary, for $NBO PLOT $END will output what is needed for visualization of all natural orbital types not just NBO in AO basis. Indeed, PNBO (prenormalized NBO) may be more useful for visualization of overlaps than NBO. PLOT causes the generation of these file types:
molname.41 AO density matrix
molname.46 Basis label file
Regardless of what ESS NBO 5 is imbedded in or what ESS GenNBO 5 derives its input from the format of these files is unchanged, so this is a completely uniform input for natural orbital visualization. Also, these inputs make a visualization tool fully functional for all natural orbital types. Also, all these types of natural orbitals (including PNBO) are available in a machine oriented format rather than having to be picked out of various places in a somewhat more human oriented format of an ESS log or output file or nbo output file (.nbo). Also, the formats .31 .. .41, .46 could be expected to be more stable across NBO updates than that of log files that might see modification in the interests of human readability.
The .31 file must be present for visualization as it contains such good things as atomic number. The different types of natural orbitals are useful depending on what is being investigated. The thing to do would be to supply a directoryname/molname and let all the .31-.41,.46 be loaded with a single command, then specify what basis you wanted to be working in at that moment.
I will attach a zipped sample file of each file type, though the .33 file might have to be compressed with another method, say zipx, to get under the upload limit.
I can imagine Jmol quite rapidly becoming the best tool to visualize natural orbitals.