so it seems that you are aware of the limitations of MIDI (only a single pitch-bend parameter per channel), but would like to shape the pitchbend value individually for each note.
assume you have created 128 tables (named "bend1" .. "bend128"), each 16383 values long, containing a lookup table to convert between input pitchbend values (-8192..+8192) and output pitchbend values (from -1..+1), then you could do something like:
[notein 1]
| |
[t b f] |
| +---------+ |
| | |
| [bendin 1] | |
| | | |
[f ] [t f f] |
| | | |
[+ 8192] [set array$1( | |
| ______________+ | |
|/ / |
[tabread] / |
| / |
[* 12] / |
| / |
[+ ] |
| |
[noteout 1]
this will modify each MIDI-note based on an individually shaped pitch-bend value.
the main drawback is, that [noteout] will send MIDI-notes (which due to the limitations of MIDI can only be integer values), so the output will be quantized to semi-tones :-(
however, if you are doing the synth within Pd, then you can use fractal note-values: just send them to [mtof] to get a frequency from the MIDI-notenumber and send it to your favourite [osc~].