图片测试

风景图

文字测试

https://xiaofan-blog-site.oss-cn-beijing.aliyuncs.com/img/20260708173154639.png

代码测试

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

import sys
from datetime import timedelta
from pathlib import Path
from faster_whisper import WhisperModel

# ============ 配置变量（按需修改）============
MODEL_SIZE = "small"          # 模型大小: tiny/base/small/medium/large-v3
LANGUAGE  = "zh"              # 语言: zh / ja / en / None(自动检测)
DEVICE    = "auto"            # "auto" / "cpu" / "cuda"
COMPUTE_TYPE = "default"      # "default" / "float16" / "int8" / "int8_float16"

INPUT_DIR  = r"G:\audio\mp3"     # MP3 文件所在目录
OUTPUT_DIR = r"G:\audio\md"      # 输出 Markdown 文件目录
# ===========================================

def _format_timestamp(seconds: float) -> str:
    """把秒数格式化为 HH:MM:SS 形式"""
    td = timedelta(seconds=seconds)
    total_seconds = int(td.total_seconds())
    hours, remainder = divmod(total_seconds, 3600)
    minutes, secs = divmod(remainder, 60)
    return f"{hours:02d}:{minutes:02d}:{secs:02d}"


def transcribe_mp3(model: WhisperModel, mp3_path: Path, output_dir: Path) -> bool:
    """转写单个 mp3 文件，输出 .md (Markdown) 文件"""
    try:
        print(f"  正在转写: {mp3_path.name} ... ", end="")
        segments, info = model.transcribe(
            str(mp3_path),
            language=LANGUAGE,
            beam_size=5,
        )

        md_path = output_dir / f"{mp3_path.stem}.md"
        duration = round(info.duration, 2)
        detected_lang = info.language

        with open(md_path, "w", encoding="utf-8") as f:
            # 文件头：标题与元信息
            f.write(f"# {mp3_path.stem}\n\n")
            f.write(f"- **文件名**: `{mp3_path.name}`\n")
            f.write(f"- **语言**: {detected_lang}\n")
            f.write(f"- **时长**: {duration}s（{_format_timestamp(duration)}）\n\n")
            f.write("---\n\n")
            # 正文：带时间戳的转写段落
            for segment in segments:
                start_ts = _format_timestamp(segment.start)
                text = segment.text.strip()
                if not text:
                    continue
                f.write(f"**[{start_ts}]** {text}\n\n")

        print(f"完成 (音频 {duration}s)")
        return True
    except Exception as e:
        print(f"失败: {e}")
        return False


def main():
    input_dir = Path(INPUT_DIR)
    output_dir = Path(OUTPUT_DIR)

    # 检查输入目录
    if not input_dir.is_dir():
        print(f"错误: 输入目录不存在 -> {input_dir}")
        sys.exit(1)

    # 创建输出目录
    output_dir.mkdir(parents=True, exist_ok=True)

    # 收集所有 mp3 文件
    mp3_files = sorted(input_dir.glob("*.mp3"))
    if not mp3_files:
        print(f"在 {input_dir} 中未找到任何 .mp3 文件")
        return

    print(f"找到 {len(mp3_files)} 个 MP3 文件")
    print(f"正在加载模型: {MODEL_SIZE} ...")

    # 加载模型
    model = WhisperModel(
        MODEL_SIZE,
        device=DEVICE,
        compute_type=COMPUTE_TYPE,
    )

    print("模型加载完成，开始转写...\n")

    # 逐个转写
    success = 0
    fail = 0
    for mp3 in mp3_files:
        ok = transcribe_mp3(model, mp3, output_dir)
        if ok:
            success += 1
        else:
            fail += 1

    # 汇总
    print(f"\n{'='*40}")
    print(f"全部完成! 成功: {success}, 失败: {fail}")
    print(f"输出目录: {output_dir}")


if __name__ == "__main__":
    main()