Hey John, how do I go about uninstalling numen? I just used 'sudo rm -rf numen' to delete the directory but I noticed the models are left behind in /usr/local/share so I did 'sudo rm -rf /usr/local/share/vosk-models' as well. If I don't delete the vosk-models directory then when I install again I end up with multiple models. After deleting vosk-models, I was still able to use the same running process of numen as usual. So I also deleted ~/.local/share/numen. After this and trying to use transcription, numen finally crashed. Is there anything else I should delete to do a clean reinstall? 

I'm still trying to fix the misspellings and typos so I'm thinking a clean reinstall may help but I may try reinstalling some pulseaudio and alsa packages as well or reaching out to the vosk team.

Hi sublime, sucks you're still getting mistypes.
The install scripts should overwrite all the old files anyway but worth a shot.
(maybe try the rest of the email first)

These should be all the paths that could possibly matter and some:

/usr/local/bin/numen
/usr/local/bin/displaying
/usr/share/man/man1/numen.1
/usr/libexec/numen
/etc/numen
/usr/local/share/vosk-models
/usr/bin/dotool
/usr/bin/dotoolc
/usr/bin/dotoold
/usr/share/vosk-models
/usr/bin/numen
/usr/bin/displaying

(You shouldn't have the last three which is where they'd be installed by a
package manager.)

I improved the install scripts a little, and should have fixed the multiple
models issue, so `git pull` before reinstalling.

I doubt it's a sound issue as the speech recognition only knows real words
and can't output gobbledygook.

Running `pgrep numen` should only output one number (just to make sure you've
only got one instance of numen running).

dotool is the program that simulates typing.  If the mistypes are only in
specific programs, dotool may be typing too fast for them, and it would be
good to know which programs.  I'm even getting mistypes when I keep trying
something like this:

	echo type this is a really really really really long sentence | dotool

but numen sets dotool's typedelay (see `dotool -h`).

You could try upping the delay by editing this line in handlers/kernel:
	alias dotool='{ print "keydelay 8" "typedelay 4"; cat; } | dotool'
to something like this:
	alias dotool='{ print "keydelay 8" "typedelay 12"; cat; } | dotool'
and reinstalling and restarting (you should notice slower typing of transcripts).

Hey John, I appreciate you getting back to me on this. I would have responded sooner but didn't wanna speak too soon again. Changing the type delay like you said seemed to correct the miss types and misspellings. I haven't ran into any more issues but will let you know if I'm experiencing problems again. If I do run into problems again I'll take note of which programs. Since that seemed to correct the issue I didn't delete the directories that may matter but may give that a try later if needed. By the way, I only have one instance of numen running. Much better experience now so thank you :).

I am still having a little bit of trouble but maybe some of this is normal with vosk. 
- Sometimes text is getting typed after saying a command or unofficial command such as ctrl+left/right to move about words. Maybe I need to adjust the key delay like I did with type delay?
- Words are sometimes being typed before I say anything after saying scribe or transcribe. Some of your usual 'the', 'in the', 'and', etc. but also things like 'made of'. I know you said some of this is normal but I've never experienced it before saying anything. I may need to tweak my microphone or something else still.