Not works on ubuntu

[david-shiko]

Linux david-comp 6.14.0-37-generic #37~24.04.1-Ubuntu SMP PREEMPT_DYNAMIC Thu Nov 20 10:25:38 UTC 2 x86_64 x86_64 x86_64 GNU/Linux
wayland
david@david-comp:~/nerd-dictation$ pip show vosk | grep Version
Version: 0.3.45

david@david-comp:~/nerd-dictation$ arecord -d 5 -r 16000 -f S16_LE -c 1 -t wav health_check.wav && file health_check.wav && soxi health_check.wav
Recording WAVE 'health_check.wav' : Signed 16 bit Little Endian, Rate 16000 Hz, Mono
health_check.wav: RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, mono 16000 Hz

Input File     : 'health_check.wav'
Channels       : 1
Sample Rate    : 16000
Precision      : 16-bit
Duration       : 00:00:05.00 = 80000 samples ~ 375 CDDA sectors
File Size      : 160k
Bit Rate       : 256k
Sample Encoding: 16-bit Signed Integer PCM

git log (nerd doesnt not have --version key).
commit 41f372789c640e01bb6650339a78312661530843 (HEAD -> main, origin/main, origin/HEAD)
Author: Campbell Barton <ideasman42@gmail.com>
Date:   Fri Oct 10 21:23:16 2025 +1100

    Cleanup: correct typo in function name

david@david-comp:~/nerd-dictation$ ~/.local/share/pipx/venvs/vosk/bin/python3 ~/nerd-dictation/nerd-dictation begin --verbose 3
Loading model...
Model loaded.

#Nothing...

and this one short script works fine

import os
from vosk import Model, KaldiRecognizer

model_path = "/home/david/.config/nerd-dictation/model"

model = Model(model_path)
rec = KaldiRecognizer(model, 16000)

with open("test_16k.raw", "rb") as f:
    while True:
        data = f.read(4000)
        if rec.AcceptWaveform(data):
            print("Tmp result:", rec.Result())
    
    print(rec.FinalResult())