http://forum.pdpatchrepo.info/topic/9799/gui-tab-like-question

https://www.iannix.org/

[bob~] ?

you can run a pd subprocess using the [pd~] object. take a look at the help file.

"I'm also running audio at the same time but the problem persists even if audio is not playing."

even so you might want to run separate instances of pd as suggested here.

http://forum.pdpatchrepo.info/topic/10026/syncing-issues-with-audio-video-have-searched-github-to-no-avail/4

http://bitsynthesis.com/pd/
?
http://bitsynthesis.com/assets/downloads/pd/bb_sync2gui.zip

you`re correct. posted for reference.
this topic is beyond me but more advanced people might weigh in.

a vocoder might make isolating the elements you are interested in easier.
rjlib has a patch called e_forcepitch-help.pd (also a_onset-help.pd)

or maybe look at synthesis / how to generate might give you ideas as to how to subtract.
https://github.com/proteusvacuum/KlattSynth
text-to-speech synthesis with libflite
moocow/flite

good luck

p.s.
http://forum.pdpatchrepo.info/topic/1844/sample-slicer-with-user-selectable-slices/14
there is a patch here called bonks_slices.pd
its not really what you want but maybe you could modify the idea to isolate words and then split a word into parts.

http://forum.pdpatchrepo.info/topic/10302/openpanel-and-readsf-play-audio-file-detuned-and-slowed-down

http://forum.pdpatchrepo.info/topic/9564/is-this-msp-patch-amplitude-gate-possible-in-pd-extended/4