Kr08 commited on
Commit
7dccca6
·
verified ·
1 Parent(s): 605a998

Update audio_processing.py

Browse files
Files changed (1) hide show
  1. audio_processing.py +5 -0
audio_processing.py CHANGED
@@ -3,6 +3,7 @@ import whisper
3
  import numpy as np
4
  import torchaudio as ta
5
  import gradio as gr
 
6
  from model_utils import get_processor, get_model, get_whisper_model_small, get_device
7
  from config import SAMPLING_RATE, CHUNK_LENGTH_S
8
  import subprocess
@@ -10,12 +11,14 @@ import subprocess
10
  import subprocess
11
  import torchaudio as ta
12
 
 
13
  def resample_with_ffmpeg(input_file, output_file, target_sr=16000):
14
  command = [
15
  'ffmpeg', '-i', input_file, '-ar', str(target_sr), output_file
16
  ]
17
  subprocess.run(command, check=True)
18
 
 
19
  def detect_language(audio):
20
  whisper_model = get_whisper_model_small()
21
 
@@ -50,6 +53,7 @@ def detect_language(audio):
50
  return detected_lang
51
 
52
 
 
53
  def process_long_audio(audio, task="transcribe", language=None):
54
  if audio[0] != SAMPLING_RATE:
55
  # Save the input audio to a file for ffmpeg processing
@@ -106,6 +110,7 @@ def process_long_audio(audio, task="transcribe", language=None):
106
  return " ".join(results)
107
 
108
 
 
109
  def process_audio(audio):
110
  if audio is None:
111
  return "No file uploaded", "", ""
 
3
  import numpy as np
4
  import torchaudio as ta
5
  import gradio as gr
6
+ import spaces
7
  from model_utils import get_processor, get_model, get_whisper_model_small, get_device
8
  from config import SAMPLING_RATE, CHUNK_LENGTH_S
9
  import subprocess
 
11
  import subprocess
12
  import torchaudio as ta
13
 
14
+
15
  def resample_with_ffmpeg(input_file, output_file, target_sr=16000):
16
  command = [
17
  'ffmpeg', '-i', input_file, '-ar', str(target_sr), output_file
18
  ]
19
  subprocess.run(command, check=True)
20
 
21
+ +@spaces.GPU
22
  def detect_language(audio):
23
  whisper_model = get_whisper_model_small()
24
 
 
53
  return detected_lang
54
 
55
 
56
+ +@spaces.GPU
57
  def process_long_audio(audio, task="transcribe", language=None):
58
  if audio[0] != SAMPLING_RATE:
59
  # Save the input audio to a file for ffmpeg processing
 
110
  return " ".join(results)
111
 
112
 
113
+ +@spaces.GPU
114
  def process_audio(audio):
115
  if audio is None:
116
  return "No file uploaded", "", ""