generated from thinkode/modelRepository
32 lines
962 B
Python
32 lines
962 B
Python
# from faster_whisper import WhisperModel
|
|
import whisper
|
|
import argparse
|
|
from log import log_step
|
|
from genericpath import exists
|
|
|
|
try:
|
|
# Create the parser
|
|
parser = argparse.ArgumentParser(description="dubstudio audio transcription tool (v1.0) - V. BOULANGER - 2025")
|
|
|
|
# Options definition
|
|
parser.add_argument('-a', '--audio', help='Audio file to process', default='audio.wav')
|
|
parser.add_argument('-m', '--model', help="Transcription model to use for processing the audio", default='small')
|
|
|
|
# Options analyzing
|
|
args = parser.parse_args()
|
|
|
|
log_step('init', 100, {
|
|
"audioFile": args.audio,
|
|
"model": args.model,
|
|
})
|
|
|
|
model = whisper.load_model(args.model)
|
|
|
|
result = model.transcribe(args.audio, word_timestamps=True, language=None, verbose=False)
|
|
|
|
except KeyboardInterrupt:
|
|
# Delete the output folder
|
|
log_step("exit", 100, "program exit")
|
|
|
|
except Exception as e:
|
|
log_step("error", 100, str(e)) |