Spaces:
Runtime error
Runtime error
File size: 1,987 Bytes
3e533d7 77070f9 3e533d7 e698260 3179a73 3e533d7 e698260 3179a73 8caee60 18e34db 3e533d7 3130060 3e533d7 d15328c cb85517 ed6e5d8 581b947 ed6e5d8 bdec318 3e533d7 ed6e5d8 3e533d7 581b947 3e533d7 581b947 3e533d7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 |
from pyannote.audio import Pipeline
from pydub import AudioSegment
from file_name import *
import torch
import json
import gc
import os
gc.collect()
torch.cuda.empty_cache()
hugging_face_token = os.environ["HUGGING_FACE_TOKEN"]
pipeline = Pipeline.from_pretrained(
'pyannote/speaker-diarization', use_auth_token=hugging_face_token)
device = torch.device("cuda")
pipeline.to(device)
def start_diarization(input_file):
diarization = pipeline(input_file)
sample_groups = []
speaker_groups = {}
for turn, _, speaker in diarization.itertracks(yield_label=True):
if (speaker not in sample_groups):
sample_groups.append(str(speaker))
suffix = 1
file_name = f"{speaker}-{suffix}"
while file_name in speaker_groups:
suffix += 1
file_name = f"{speaker}-{suffix}"
speaker_groups[file_name] = [turn.start, turn.end]
print(f"speaker_groups {file_name}: {speaker_groups[file_name]}")
print(
f"start={turn.start:.3f}s stop={turn.end:.3f}s speaker_{speaker}")
save_groups_json(input_file, sample_groups, speaker_groups)
audio_segmentation(input_file, speaker_groups)
print(str(speaker_groups))
return str(speaker_groups)
def audio_segmentation(input_file, speaker_groups_dict):
audioSegment = AudioSegment.from_wav(input_file)
for speaker in speaker_groups_dict:
time = speaker_groups_dict[speaker]
audioSegment[time[0]*1000: time[1] *
1000].export(f"{speaker}.wav", format='wav')
print(f"group {speaker}: {time[0]*1000}--{time[1]*1000}")
def save_groups_json(input_file, sample_groups_list: list, speaker_groups_dict: dict):
with open(sample_groups_json, "w", encoding="utf-8") as json_file_sample:
json.dump(sample_groups_list, json_file_sample)
with open(speaker_groups_json, "w", encoding="utf-8") as json_file_speaker:
json.dump(speaker_groups_dict, json_file_speaker)
|