I have prepared a Speech to Text system using using Kaldi and vosk.
Ive setup the directories and I am using vosk to give transcription of audio files.
The pipeline is that I run bash script which basically takes in audio file name (without extension), breaks it into chunks which are saved in a folder in the same directory. Then it runs the transcription script (vosk API with kaldi model) on each chunk and get a transcript in text file and then all chunk text are saved into one text-file.
The bash code is as follows: (#usage is $0 <audiofilename-without-extension)
#! bin/bash
af=$1
afe= $af + ".wav"
python3 chunker.py "$af"
for file in ${af}/*.wav;
do
python3 test_ffmpeg.py "$file" >> ${file}.txt
done
for f in ${af}/*.txt;
do
echo -e $(cat "$f") '\n' >> ${af}.txt
done
The output format I get is this:
{
"partial" : "assalamualaikum c p l c call karney ka shukria operator 13 baat kar"
}
{
"partial" : "assalamualaikum c p l c call karney ka shukria operator 13 baat kar"
}
{
"text" : "assalamualaikum c p l c call karney ka shukria operator 13 baat kar"
}
What I want in my output is the {"text": ""} part only, that too without the {"text":""}. Can anyone guide me how to achieve this output?
The other scripts mentioned in the bash file are as follows:
test_ffmpeg.py is from vosk-api example scripts which is as follows:
#!/usr/bin/env python3
from vosk import Model, KaldiRecognizer, SetLogLevel
import sys
import os
import wave
import subprocess
import srt
import json
import datetime
SetLogLevel(0)
sample_rate=16000
model = Model("..")
rec = KaldiRecognizer(model, sample_rate)
process = subprocess.Popen(['ffmpeg', '-loglevel', 'quiet', '-i',
sys.argv[1],
'-ar', str(sample_rate) , '-ac', '1', '-f', 's16le', '-'],
stdout=subprocess.PIPE)
while True:
data = process.stdout.read(4000)
if len(data) == 0:
break
if rec.AcceptWaveform(data):
print(rec.Result())
else:
print(rec.PartialResult())
print(rec.FinalResult())
chunker.py uses the $1 audiofilename and breaks it into chunks in a folder named as $1 variable. So if the wav file name is call21.wav it will create a folder called call21 and save all chunk files as chunk1.wav chunk2.wav and so on
import speech_recognition as sr
import os
import pyaudio
from pydub import AudioSegment
from pydub.silence import split_on_silence
from vosk import Model, KaldiRecognizer, SetLogLevel
import wave
import sys
import subprocess
fname = sys.argv[1] #enter name without extension
wav = ".wav"
txt = ".txt"
transcript = fname + txt
audiofilename = fname + wav
sample_rate=16000
SetLogLevel(-1)
path = audiofilename
#recognizer.SetWords(True)
#recognizer.SetPartialWords(True)
# open the audio file using pydub
sound = AudioSegment.from_wav(path)
# split audio sound where silence is 700 miliseconds or more and get chunks
chunks = split_on_silence(sound,
# experiment with this value for your target audio file
min_silence_len = 1000,
# adjust this per requirement
silence_thresh = sound.dBFS-16,
# keep the silence for 1 second, adjustable as well
keep_silence=2000,
)
folder_name = fname
# create a directory to store the audio chunks
if not os.path.isdir(folder_name):
os.mkdir(folder_name)
whole_text = ""
# process each chunk
for i, audio_chunk in enumerate(chunks, start=1):
# export audio chunk and save it in
# the `folder_name` directory.
chunk_filename = os.path.join(folder_name, f"chunk{i}.wav")
audio_chunk.export(chunk_filename, format="wav")
if __name__ == '__main__':
import sys
path = audiofilename
#path = sys.argv[1]```
