Adding the word "the" before and after saying anything or if I pause it just types "the" again

[bobhenkel]

Any help would be appreciated, here's my setup and a sample. I verbally said "the" 0 times yet it shows up automatically. My thoughts are unless I say the word "the" it shouldn't just be prefixed or suffixed on what I say.

❯ nerd-dictation begin --pulse-device-name='alsa_input.usb-046d_HD_Webcam_C615_E47E8AA0-00.mono-fallback' --vosk-model-dir /home/bob/vosk-model-en-us-0.22 --simulate-input-tool YDOTOOL
the after waiting a few seconds about it types that automatically and then if I pause again waiting a few seconds about it types it automatically again and then if I pause again for a moment here the edited it again the

Thanks!

[ideasman42]

Hi, these kinds of issues are really out of the scope of this project to solve,
this would be todo with vosk & the model your using.
It's possibly related to your audio setup if there is any additional noise - although it's hard to say.

Personally I don't have this problem - using the vosk-model-en-us-0.42-gigaspeech model (with suspend/resume on Linux as it's a little slow to load).

• What model are you using?
• Can you try a different model (such as the one mentioned above)?

[cboecking]

Note that I experienced the same thing. I just found and installed nerd-dictation for the first time today. The system adds about 15 'the's per minute to the prompt even when system mic is completely disabled. I am also using the vosk-model-en-us-0.22 model.

geeky details:

• i have incus installed, and have a container per model (small, reg, giga)
• i have all three models running and paused (ctrl+z)
• the regular (vosk-model-en-us-0.22) is the only model exhibiting this behavior

more details:

• installed and ran on debian 12 using: https://github.com/chuboe/chuboe-system-configurator/blob/main/nerd-dictation/nerd.md

I hope this helps!

[Mayurifag]

Same model and same problem.

Did anyone solved it already?

I also might have problem with ydotool because system might easily go crazy when I use nerd-dictation. I am on latest cachyos/kde/wayland setup.

Update: nevermind, tried another model, vosk-model-en-us-0.42-gigaspeech - worked MUCH better and WITHOUT mentioned problem, even though it was kinda slow for me for first usage. Locked into testing.

vosk-model-en-us-0.22 is bugged.