dotfiles/.local/bin/scripts/vosk

43 lines
1.2 KiB
Plaintext
Raw Normal View History

2020-10-29 10:45:07 +00:00
#!/usr/bin/env python3
from vosk import Model, KaldiRecognizer, SetLogLevel
import sys
import json
import os
import wave
import subprocess
SetLogLevel(0)
modelBase = os.path.join(os.getenv('XDG_DATA_HOME', os.environ['HOME']), 'vosk/models')
2020-10-29 10:45:07 +00:00
modelPath = os.path.join(modelBase, sys.argv[2] if len(sys.argv) > 2 else 'daanzu')
if not os.path.exists(modelPath):
print (f"Please download a model from https://alphacephei.com/vosk/models and unpack it to {modelPath}.", file=sys.stderr)
exit (1)
sample_rate=16000
model = Model(modelPath)
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)
def getText(result):
res = json.loads(result)
if not 'result' in res:
return res['text']
sec = res['result'][0]['start']
time = "%02d:%02d" % (sec // 60, sec % 60)
return "%s %s" % (time, res['text'])
2020-10-29 10:45:07 +00:00
while True:
data = process.stdout.read(4000)
if len(data) == 0:
break
if rec.AcceptWaveform(data):
print(getText(rec.Result()))
2020-10-29 10:45:07 +00:00
print(getText(rec.FinalResult()))