File size: 2,192 Bytes
e2d8d82
931df81
 
 
 
 
 
 
e2d8d82
931df81
 
 
cb85517
32e4ded
cb85517
931df81
 
 
3130060
931df81
3130060
32e4ded
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
931df81
 
 
 
 
 
 
 
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
60
61
62
from faster_whisper import WhisperModel
import torch
import gc
import json

gc.collect()
torch.cuda.empty_cache()

model = WhisperModel("medium", device="cuda", compute_type="int8_float16")


def start_transcribe(progress):
    _, speaker_groups = load_groups_json()

    for speaker, _ in zip(speaker_groups, progress.tqdm(speaker_groups, desc="Processing diarization")):
        # Transcribe and save temp file
        audiof = f"{speaker}.wav"
        print(f"Loading {audiof}")
        segments, _ = model.transcribe(
            audio=audiof, language='id', word_timestamps=True)
        segments_list = list(segments)

        text_list_to_print = []
        for segment in segments_list:
            start = timeStr(segment['start'])
            end = timeStr(segment['end'])
            name = str(speaker)[:10]
            text = segment["text"]
            subtitle_txt = f"{len(subtitle_txt) + 1}\n{start} --> {end}\n[{name}] {text}\n\n"
            # Appending subtitle txt for each segment
            with open("subtitle.srt", "a") as file:
                file.writelines(subtitle_txt)
            # Appending text for each segment to print
            text_list_to_print.append(text)

        # Print full text for each speaker turn
        text = "\n".join(text_list_to_print)
        print(text)

        # Create transcribe per speaker
        with open(f"{speaker}.json", "w") as text_file:
            json.dump(segments_list, text_file, indent=4)
        # Append to complete transcribe file
        with open("transcribe.txt", "a") as file:
            file.write(f"[{name}] {text}\n")

    return ["subtitle.srt", "transcribe.txt"]


def timeStr(t):
    return '{0:02d}:{1:02d}:{2:06.2f}'.format(round(t // 3600),
                                              round(t % 3600 // 60),
                                              t % 60)


def load_groups_json():
    with open("sample_groups.json", "r") as json_file_sample:
        sample_groups_list: list = json.load(json_file_sample)
    with open("speaker_groups.json", "r") as json_file_speaker:
        speaker_groups_dict: dict = json.load(json_file_speaker)
    return sample_groups_list, speaker_groups_dict