Spaces:
Sleeping
Sleeping
Delete whisper_processor (1).py
Browse files- whisper_processor (1).py +0 -71
whisper_processor (1).py
DELETED
@@ -1,71 +0,0 @@
|
|
1 |
-
import subprocess
|
2 |
-
import sys
|
3 |
-
import os
|
4 |
-
from pydub import AudioSegment
|
5 |
-
import io
|
6 |
-
import wave
|
7 |
-
|
8 |
-
def process_audio(wav_file, model_name, lang):
|
9 |
-
"""
|
10 |
-
Processes an audio file using a specified model and returns the processed string.
|
11 |
-
Resamples the file to 16 kHz in-memory if needed.
|
12 |
-
|
13 |
-
:param wav_file: Path to the WAV file
|
14 |
-
:param model_name: Name of the model to use
|
15 |
-
:return: Processed string output from the audio processing
|
16 |
-
:raises: Exception if an error occurs during processing
|
17 |
-
"""
|
18 |
-
model = f"models/ggml-{model_name}.bin"
|
19 |
-
|
20 |
-
# Check if the model file exists
|
21 |
-
if not os.path.exists(model):
|
22 |
-
raise FileNotFoundError(f"Model file not found: {model} \n\nDownload a model with this command:\n\n> bash ./models/download-ggml-model.sh {model_name}\n\n")
|
23 |
-
|
24 |
-
if not os.path.exists(wav_file):
|
25 |
-
raise FileNotFoundError(f"WAV file not found: {wav_file}")
|
26 |
-
|
27 |
-
# Load and resample the audio to 16 kHz if necessary
|
28 |
-
audio = AudioSegment.from_wav(wav_file)
|
29 |
-
if audio.frame_rate != 16000:
|
30 |
-
print(f"Resampling {wav_file} to 16 kHz...")
|
31 |
-
audio = audio.set_frame_rate(16000)
|
32 |
-
|
33 |
-
# Export the audio to an in-memory bytes buffer
|
34 |
-
audio_buffer = io.BytesIO()
|
35 |
-
audio.export(audio_buffer, format="wav")
|
36 |
-
audio_buffer.seek(0)
|
37 |
-
|
38 |
-
# Prepare the audio for the subprocess by writing it to a temporary file
|
39 |
-
with wave.open(audio_buffer, 'rb') as wf:
|
40 |
-
# Prepare a temporary file in-memory for subprocess
|
41 |
-
temp_file = io.BytesIO()
|
42 |
-
with wave.open(temp_file, 'wb') as temp_wav:
|
43 |
-
temp_wav.setnchannels(wf.getnchannels())
|
44 |
-
temp_wav.setsampwidth(wf.getsampwidth())
|
45 |
-
temp_wav.setframerate(16000)
|
46 |
-
temp_wav.writeframes(wf.readframes(wf.getnframes()))
|
47 |
-
temp_file.seek(0)
|
48 |
-
|
49 |
-
# Write temp_file to disk if needed, or pass to subprocess directly if possible.
|
50 |
-
# Here we assume it's passed to subprocess via its filename as before
|
51 |
-
temp_wav_path = '/tmp/temp_audio.wav'
|
52 |
-
with open(temp_wav_path, 'wb') as f:
|
53 |
-
f.write(temp_file.read())
|
54 |
-
|
55 |
-
# Run the processing using the temporary WAV file
|
56 |
-
full_command = f"./main -m {model} -f {temp_wav_path} -l {lang} -np -nt"
|
57 |
-
|
58 |
-
# Execute the command
|
59 |
-
process = subprocess.Popen(full_command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
|
60 |
-
|
61 |
-
# Get the output and error (if any)
|
62 |
-
output, error = process.communicate()
|
63 |
-
|
64 |
-
if error:
|
65 |
-
raise Exception(f"Error processing audio: {error.decode('utf-8')}")
|
66 |
-
|
67 |
-
# Process and return the output string
|
68 |
-
decoded_str = output.decode('utf-8').strip()
|
69 |
-
processed_str = decoded_str.replace('[BLANK_AUDIO]', '').strip()
|
70 |
-
|
71 |
-
return processed_str
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|