File size: 8,270 Bytes
0d6f640
 
 
 
 
 
 
 
 
 
3c11817
 
0d6f640
 
 
 
 
 
 
 
 
 
 
3c11817
 
0d6f640
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3c11817
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0d6f640
 
 
 
 
 
3c11817
 
0d6f640
 
3c11817
0d6f640
 
 
 
 
 
3c11817
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0d6f640
 
 
 
 
 
 
3c11817
0d6f640
 
 
 
 
 
 
 
3c11817
0d6f640
 
 
 
 
 
 
 
3c11817
 
 
 
 
 
0d6f640
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
"""Utility functions for extracting audio, transcribing and merging subtitles."""

from __future__ import annotations

import logging
import os
import subprocess
from dataclasses import dataclass
from typing import List, Optional

from pydub import AudioSegment

# MoviePy is an optional dependency used when extracting audio. It is imported
# lazily to avoid issues when running in environments where it is not
# available (for instance during unit tests).

try:
    from faster_whisper import WhisperModel
except ImportError:  # pragma: no cover - optional dependency
    WhisperModel = None

logging.basicConfig(level=logging.DEBUG, format="%(asctime)s - %(levelname)s - %(message)s")

MAX_OPENAI_AUDIO_SIZE = 25 * 1024 * 1024  # 25 MB


def format_timestamp(seconds: float) -> str:
    """Return timestamp in SRT format."""
    h = int(seconds // 3600)
    m = int((seconds % 3600) // 60)
    s = int(seconds % 60)
    ms = int((seconds - int(seconds)) * 1000)
    return f"{h:02}:{m:02}:{s:02},{ms:03}"


def extract_audio(video_path: str, output_dir: str) -> str:
    """Extract audio from *video_path* and return the audio file path."""
    if not os.path.exists(video_path):
        raise FileNotFoundError(video_path)
    os.makedirs(output_dir, exist_ok=True)
    base_name = os.path.splitext(os.path.basename(video_path))[0]
    audio_path = os.path.join(output_dir, f"{base_name}.wav")
    # Import here so tests that do not require MoviePy can run without the
    # dependency installed.
    from moviepy.editor import VideoFileClip

    clip = VideoFileClip(video_path)
    clip.audio.write_audiofile(audio_path, logger=None)
    clip.close()
    return audio_path


@dataclass
class SubtitleLine:
    start: float
    end: float
    text: str


def _segments_to_srt(segments: List[SubtitleLine]) -> str:
    lines = []
    for idx, seg in enumerate(segments, 1):
        lines.append(str(idx))
        lines.append(f"{format_timestamp(seg.start)} --> {format_timestamp(seg.end)}")
        lines.append(seg.text.strip())
        lines.append("")
    return "\n".join(lines)


def _export_and_transcribe_segment(seg, idx, audio_path, openai, words_per_sub, time_offset):
    """Esporta un segmento in MP3, verifica la dimensione e lo suddivide ricorsivamente se necessario."""
    import tempfile
    segment_list = []
    txt_list = []
    with tempfile.NamedTemporaryFile(suffix=f"_part{idx}.mp3", delete=False) as temp_file:
        seg.export(temp_file.name, format="mp3")
        temp_size = os.path.getsize(temp_file.name)
        logging.debug(f"Segmento {idx}: dimensione {temp_size} byte (MP3)")
        if temp_size > MAX_OPENAI_AUDIO_SIZE:
            # Suddividi ulteriormente il segmento
            logging.info(f"Segmento {idx} ancora troppo grande, suddivisione ricorsiva...")
            duration_ms = len(seg)
            mid = duration_ms // 2
            seg1 = seg[:mid]
            seg2 = seg[mid:]
            # Ricorsione su ciascuna metà
            segs1, txts1 = _export_and_transcribe_segment(seg1, f"{idx}a", audio_path, openai, words_per_sub, time_offset)
            segs2, txts2 = _export_and_transcribe_segment(seg2, f"{idx}b", audio_path, openai, words_per_sub, time_offset + seg1.duration_seconds)
            segment_list.extend(segs1)
            segment_list.extend(segs2)
            txt_list.extend(txts1)
            txt_list.extend(txts2)
        else:
            with open(temp_file.name, "rb") as audio_file:
                result = openai.audio.transcriptions.create(
                    model="whisper-1",
                    file=audio_file,
                    response_format="json",
                )
            words = result.text.split()
            plain = result.text.strip()
            txt_list.append(plain)
            # Ricostruisci segmenti SRT con offset temporale
            segs = []
            start = time_offset
            step = 3.0
            for i in range(0, len(words), words_per_sub):
                end = start + step
                text = " ".join(words[i : i + words_per_sub])
                segs.append(SubtitleLine(start=start, end=end, text=text))
                start = end
            segment_list.extend(segs)
        os.remove(temp_file.name)
    return segment_list, txt_list


def transcribe_audio(
    audio_path: str,
    library: str = "faster_whisper",
    api_key: Optional[str] = None,
    model_size: str = "base",
    words_per_sub: int = 7,
) -> tuple[str, str]:
    """Transcribe *audio_path* and return (SRT content, plain text content)."""
    logging.debug(f"Starting transcription with library: {library}, audio_path: {audio_path}")

    plain_text = None
    if library == "OpenAI Whisper":
        if api_key is None:
            raise ValueError("api_key is required for OpenAI Whisper")
        import openai

        openai.api_key = api_key
        # --- Gestione file troppo grandi ---
        if os.path.getsize(audio_path) > MAX_OPENAI_AUDIO_SIZE:
            logging.info("Audio troppo grande, suddivisione in segmenti...")
            audio = AudioSegment.from_file(audio_path)
            duration_ms = len(audio)
            segment_length_ms = 20 * 60 * 1000
            segments = [audio[i : i + segment_length_ms] for i in range(0, duration_ms, segment_length_ms)]
            srt_parts = []
            txt_parts = []
            time_offset = 0.0
            for idx, seg in enumerate(segments):
                segs, txts = _export_and_transcribe_segment(seg, idx, audio_path, openai, words_per_sub, time_offset)
                srt_parts.extend(segs)
                txt_parts.extend(txts)
                time_offset += seg.duration_seconds
            segments = srt_parts
            plain_text = " ".join(txt_parts)
        else:
            with open(audio_path, "rb") as audio_file:
                result = openai.audio.transcriptions.create(
                    model="whisper-1",
                    file=audio_file,
                    response_format="json",
                )
                logging.debug(f"OpenAI API response: {result}")
                words = result.text.split()
                plain_text = result.text.strip()
                if not words:
                    logging.error("No text returned by OpenAI Whisper API.")
                    raise ValueError("No text returned by OpenAI Whisper API.")
                segments = []
                start = 0.0
                step = 3.0
                for i in range(0, len(words), words_per_sub):
                    end = start + step
                    text = " ".join(words[i : i + words_per_sub])
                    segments.append(SubtitleLine(start=start, end=end, text=text))
                    start = end
                logging.debug(f"Generated segments: {segments}")
    else:
        if WhisperModel is None:
            raise RuntimeError("faster_whisper is not installed")
        logging.debug("Using Faster Whisper for transcription...")
        model = WhisperModel(model_size)
        segs = model.transcribe(audio_path)[0]
        segments = [SubtitleLine(start=s.start, end=s.end, text=s.text) for s in segs]
        plain_text = " ".join([s.text.strip() for s in segments])
        logging.debug(f"Generated segments: {segments}")

    if not segments:
        logging.error("No segments generated during transcription.")
        raise ValueError("No segments generated during transcription.")

    srt_content = _segments_to_srt(segments)
    logging.debug(f"Generated SRT content: {srt_content}")
    return srt_content, plain_text


def save_srt(content: str, output_path: str) -> str:
    with open(output_path, "w", encoding="utf-8") as f:
        f.write(content)
    return output_path


def save_txt(content: str, output_path: str) -> str:
    with open(output_path, "w", encoding="utf-8") as f:
        f.write(content)
    return output_path


def merge_subtitles(video_path: str, srt_path: str, output_path: str) -> str:
    command = [
        "ffmpeg",
        "-y",
        "-i",
        video_path,
        "-vf",
        f"subtitles={srt_path}",
        "-c:a",
        "copy",
        "-c:v",
        "libx264",
        output_path,
    ]
    subprocess.run(command, check=True)
    return output_path