Alex Volkov commited on
Commit
f1f3adb
·
1 Parent(s): 4d8662a

Added CLI to preload model by default and to be able to specify large

Browse files
Files changed (2) hide show
  1. app.py +7 -2
  2. download.py +21 -3
app.py CHANGED
@@ -9,6 +9,7 @@ from pathlib import Path
9
  from shutil import rmtree
10
  import dotenv
11
  from utils.utils import str2bool
 
12
 
13
  dotenv.load_dotenv()
14
 
@@ -19,10 +20,14 @@ parser = argparse.ArgumentParser(
19
  formatter_class=argparse.ArgumentDefaultsHelpFormatter)
20
  parser.add_argument("--public", type=str2bool, default=False,
21
  help="Wether to share with gradio public or not")
22
-
 
 
 
23
 
24
  args = parser.parse_args().__dict__
25
- gradio_share: str = args.pop("public")
 
26
 
27
  url_input = gr.Textbox(label="Youtube/Twitter/etc video URL (supports many services)", value='https://twitter.com/starsonxh/status/1552945347194142720', lines=1, elem_id="url_input")
28
  download_status = gr.Textbox(label="Status:", value='', lines=1, elem_id="download_status")
 
9
  from shutil import rmtree
10
  import dotenv
11
  from utils.utils import str2bool
12
+ import whisper
13
 
14
  dotenv.load_dotenv()
15
 
 
20
  formatter_class=argparse.ArgumentDefaultsHelpFormatter)
21
  parser.add_argument("--public", type=str2bool, default=False,
22
  help="Wether to share with gradio public or not")
23
+ parser.add_argument("--preload", type=str2bool, default=True,
24
+ help="Should the model be preloaded on script launch. Disable for faster debug")
25
+ parser.add_argument("--model", default="medium",
26
+ choices=whisper.available_models(), help="name of the Whisper model to use")
27
 
28
  args = parser.parse_args().__dict__
29
+ gradio_share: str = args.get("public")
30
+ model_size: str = args.get("model")
31
 
32
  url_input = gr.Textbox(label="Youtube/Twitter/etc video URL (supports many services)", value='https://twitter.com/starsonxh/status/1552945347194142720', lines=1, elem_id="url_input")
33
  download_status = gr.Textbox(label="Status:", value='', lines=1, elem_id="download_status")
download.py CHANGED
@@ -9,16 +9,33 @@ from youtube_dl.utils import DownloadError
9
  import os
10
  import tempfile
11
  import json
 
12
  import whisper
13
  from whisper.tokenizer import LANGUAGES
14
-
15
  import ffmpeg
16
  from utils.subs import bake_subs
 
17
 
18
  original_dir = os.getcwd()
19
  output_dir = Path('output')
20
 
21
- def download_generator(url):
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
22
  ### Step 1 : check if video is available
23
  yield {"message": f"Checking {url} for videos"}
24
  try:
@@ -138,7 +155,8 @@ def check_download(url):
138
 
139
  def transcribe(audio):
140
  print('Starting transcribe...')
141
- model = whisper.load_model('medium')
 
142
  output = model.transcribe(audio, task="translate")
143
  output["language"] = LANGUAGES[output["language"]]
144
  # output = {"text": " And then, you throw it and bam! How do you do this when it's going this way? Throw it and bam! That's right! Is that it? Throw it and bam! That's it. That's it. That's it. That's it. That's it. That's right. Throw it and bam! No, this is... Do it like this and... Isn't this the most natural? Bam! What is it?", "segments": [{"id": 0, "seek": 0, "start": 0.0, "end": 2.46, "text": " And then, you throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 1, "seek": 0, "start": 2.46, "end": 4.86, "text": " How do you do this when it's going this way?", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 2, "seek": 0, "start": 4.86, "end": 6.36, "text": " Throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 3, "seek": 0, "start": 6.36, "end": 7.92, "text": " That's right!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 4, "seek": 0, "start": 7.92, "end": 8.42, "text": " Is that it?", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 5, "seek": 0, "start": 8.42, "end": 9.76, "text": " Throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 6, "seek": 0, "start": 9.76, "end": 10.26, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 7, "seek": 0, "start": 10.26, "end": 10.76, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 8, "seek": 0, "start": 10.76, "end": 11.26, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 9, "seek": 0, "start": 11.26, "end": 11.76, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 10, "seek": 0, "start": 11.76, "end": 12.26, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 11, "seek": 0, "start": 12.26, "end": 12.76, "text": " That's right.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 12, "seek": 0, "start": 12.76, "end": 14.6, "text": " Throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 13, "seek": 0, "start": 14.6, "end": 15.32, "text": " No, this is...", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 14, "seek": 0, "start": 15.32, "end": 16.36, "text": " Do it like this and...", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 15, "seek": 0, "start": 16.36, "end": 17.92, "text": " Isn't this the most natural?", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 16, "seek": 0, "start": 17.92, "end": 18.92, "text": " Bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 17, "seek": 1892, "start": 18.92, "end": 30.92, "text": " What is it?", "tokens": [50364, 708, 307, 309, 30, 50964], "temperature": 0.0, "avg_logprob": -0.9666390419006348, "compression_ratio": 0.5789473684210527, "no_speech_prob": 0.0033069916535168886}], "language": "korean"}
 
9
  import os
10
  import tempfile
11
  import json
12
+ import argparse
13
  import whisper
14
  from whisper.tokenizer import LANGUAGES
 
15
  import ffmpeg
16
  from utils.subs import bake_subs
17
+ from utils.utils import str2bool
18
 
19
  original_dir = os.getcwd()
20
  output_dir = Path('output')
21
 
22
+ parser = argparse.ArgumentParser(
23
+ formatter_class=argparse.ArgumentDefaultsHelpFormatter)
24
+ parser.add_argument("--preload", type=str2bool, default=True,
25
+ help="Should the model be preloaded on script launch. Disable for faster debug")
26
+ parser.add_argument("--model", default="medium",
27
+ choices=whisper.available_models(), help="name of the Whisper model to use")
28
+
29
+ args = parser.parse_args().__dict__
30
+ model_size: str = args.get("model")
31
+ preload_model: str = args.get("preload")
32
+
33
+ if preload_model:
34
+ print("Preloading model")
35
+ model = whisper.load_model(model_size)
36
+
37
+ def download_generator(url, args):
38
+
39
  ### Step 1 : check if video is available
40
  yield {"message": f"Checking {url} for videos"}
41
  try:
 
155
 
156
  def transcribe(audio):
157
  print('Starting transcribe...')
158
+ if not preload_model:
159
+ model = whisper.load_model(model_size)
160
  output = model.transcribe(audio, task="translate")
161
  output["language"] = LANGUAGES[output["language"]]
162
  # output = {"text": " And then, you throw it and bam! How do you do this when it's going this way? Throw it and bam! That's right! Is that it? Throw it and bam! That's it. That's it. That's it. That's it. That's it. That's right. Throw it and bam! No, this is... Do it like this and... Isn't this the most natural? Bam! What is it?", "segments": [{"id": 0, "seek": 0, "start": 0.0, "end": 2.46, "text": " And then, you throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 1, "seek": 0, "start": 2.46, "end": 4.86, "text": " How do you do this when it's going this way?", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 2, "seek": 0, "start": 4.86, "end": 6.36, "text": " Throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 3, "seek": 0, "start": 6.36, "end": 7.92, "text": " That's right!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 4, "seek": 0, "start": 7.92, "end": 8.42, "text": " Is that it?", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 5, "seek": 0, "start": 8.42, "end": 9.76, "text": " Throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 6, "seek": 0, "start": 9.76, "end": 10.26, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 7, "seek": 0, "start": 10.26, "end": 10.76, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 8, "seek": 0, "start": 10.76, "end": 11.26, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 9, "seek": 0, "start": 11.26, "end": 11.76, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 10, "seek": 0, "start": 11.76, "end": 12.26, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 11, "seek": 0, "start": 12.26, "end": 12.76, "text": " That's right.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 12, "seek": 0, "start": 12.76, "end": 14.6, "text": " Throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 13, "seek": 0, "start": 14.6, "end": 15.32, "text": " No, this is...", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 14, "seek": 0, "start": 15.32, "end": 16.36, "text": " Do it like this and...", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 15, "seek": 0, "start": 16.36, "end": 17.92, "text": " Isn't this the most natural?", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 16, "seek": 0, "start": 17.92, "end": 18.92, "text": " Bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 17, "seek": 1892, "start": 18.92, "end": 30.92, "text": " What is it?", "tokens": [50364, 708, 307, 309, 30, 50964], "temperature": 0.0, "avg_logprob": -0.9666390419006348, "compression_ratio": 0.5789473684210527, "no_speech_prob": 0.0033069916535168886}], "language": "korean"}