【汉化工具系列 #2】指定wave格式转换为srt格式字幕（CPU版本）

shane007

4 h% o* a6 F7 ]
本程序是在Whisper.net.Demo基础上修改的,可以批量识别wav文件，
. H& k1 T5 y% \/ ^, F 使用时需要輸入语言，源文件路径和目标文件路径
8 n) V" w% e3 l3 C; ]最后是输出srt文件
5 {' I9 A  ?/ f+ r4 H+ `% j7 L; x
) ~& e9 d' Q! K: z/ f4 I代码如下
" p% V1 i  c1 d( R2 x以下这句用多线程可以增速，否则很慢
以下这份代码，还只能使用到CPU，速度依旧比较慢，使用CPU的方法另行研究。

1. var builder = factory.CreateBuilder()
   2 g5 W7 p, k+ B. j" E# G
2. .WithLanguage(languageOption).WithSpeedUp2x().WithThreads(16);

复制代码

1. // Licensed under the MIT license: https://opensource.org/licenses/MIT
   
2. 
   3 j0 Y  K% i, i4 w0 O
3. using System;
   . X) j" h; c0 \$ P$ P6 {7 Q! Y! y# E
4. using System.Diagnostics;
   ) b7 ]9 P7 N' B; L- H7 g+ D
5. using System.IO;
   ( j8 v: S. @! m9 G6 {0 w
6. using System.Threading;
   
7. using System.Threading.Tasks;
   
8. using CommandLine;
   
9. using Whisper.net;
   9 \: g. t9 D3 Z+ h
10. using Whisper.net.Ggml;
    ! Q7 D; F+ V" g9 C# l# E
11. using Whisper.net.Wave;
    $ r+ j) _2 n# a5 J
12. 
    
13. await Parser.Default.ParseArguments<Options>(args)
    
14.     .WithParsedAsync(Demo);
    
15. 
    
16. async Task Demo(Options opt)
    
17. 
    
18. {
    
19.     if (!File.Exists(opt.ModelName))
    : x2 b1 I! y1 Y" `+ G7 |" A
20.     {
    
21.         Console.WriteLine([        DISCUZ_CODE_1        ]quot;Downloading Model {opt.ModelName}");
    & L6 Z% S) Y/ L# I* h
22.         using var modelStream = await WhisperGgmlDownloader.GetGgmlModelAsync(opt.ModelType);
    / O: E+ a! O5 G$ X. h2 ?( a4 A, E) o
23.         using var fileWriter = File.OpenWrite(opt.ModelName);
    
24.         await modelStream.CopyToAsync(fileWriter);
    3 C2 |3 Q, F% X1 L5 i
25.     }
    
26. 
    6 x, t% d% \. o; W$ b8 q7 K- f$ b
27.     switch (opt.Command)
    # l' \. U0 C' W0 `& E
28.     {
    
29.         case "lang-detect":
    
30.             LanguageIdentification(opt);
    
31.             break;
    7 C0 \1 s  k9 A: V' @9 }# U
32.         case "transcribe":
    
33.         case "translate":
    # ~. e! D. c+ d7 ?
34.             await FullDetection(opt);
    
35.             break;
    % ?! Q) ~: N; U1 H% s
36.         default:
    * |) T) Q( ^( f6 G1 L) B% T0 _
37.             Console.WriteLine("Unknown command");
    
38.             break;
    : F/ |7 {, J: k5 g7 {
39.     }
    
40. }
    5 b! F' t* O( v: [! R
41. 
    . ?. {/ `' k; G& R! f( W
42. void LanguageIdentification(Options opt)
    
43. {
    
44.     var bufferedModel = File.ReadAllBytes(opt.ModelName);
    0 @6 u! _7 ]- C' [
45. 
    7 i0 c$ O! L! l9 A2 N
46.     // Same factory can be used by multiple task to create processors.
    
47.     using var factory = WhisperFactory.FromBuffer(bufferedModel);
    6 u8 D1 w7 p" l" ~5 c
48. 
    : V- [3 q% f2 V6 j4 W
49. /*    var builder = factory.CreateBuilder()
    
50.        .WithLanguage(opt.Language);*/
    % c* a0 Z5 T  h5 G
51.     var builder = factory.CreateBuilder()
    
52.    .WithLanguage("english");
    
53.     using var processor = builder.Build();
    0 t" {$ {* N0 c( |* \. W
54. 
    + m" O! s& e5 D8 f
55.     using var fileStream = File.OpenRead(opt.FileName);
    
56. 
    
57.     var wave = new WaveParser(fileStream);
    
58. 
    
59.     var samples = wave.GetAvgSamples();
    
60. 
    3 M# Q+ }/ A9 y! H5 t- L9 Z0 F
61.     var language = processor.DetectLanguage(samples, speedUp: true);
    
62.     Console.WriteLine("Language is " + language);
    / [0 C4 R" A# i" v6 B
63. }
    , z+ n5 C+ J: [* `5 a  |! d  w
64. 
    
65. async Task FullDetection(Options opt)
    
66. {
    7 ]5 a" d  F: m! h
67.     // Same factory can be used by multiple task to create processors.
    ) t8 s1 E& J' x1 v6 n  s# {
68.     using var factory = WhisperFactory.FromPath(opt.ModelName);
    & {0 C. b5 j  ?' {# f- U
69. 
    . B' j2 Q/ p) G. ]" a
70.     //   var builder = factory.CreateBuilder().WithLanguage(opt.Language);
    9 I3 p# {" k5 ]. ^5 L
71.     Console.WriteLine("请输入语言选项（例如：english，chinese, japanese等）：");
    
72.     string languageOption = Console.ReadLine();
    
73.     var builder = factory.CreateBuilder()
    
74.     .WithLanguage(languageOption).WithSpeedUp2x().WithThreads(16);
    ( c; x* t' F# P9 r& m; y& [+ ]
75. 
    9 @6 }' t% R) R3 L7 F
76.     if (opt.Command == "translate")
    & [5 I% F2 u3 P% t
77.     {
    
78.         builder.WithTranslate();
    3 ^; q7 H: _# l) u
79.     }
    
80. 
    
81.     WhisperProcessor processor = builder.Build();
    
82.    
    . T# d( V. n2 b
83.     Console.WriteLine("请输入wave源文件目录：");
    
84.     string sourceDirectory = Console.ReadLine();
    7 C3 @4 z! x( h$ t
85. 
    
86.     Console.WriteLine("请输入目标文件目录：");
    $ y7 P; F- d+ K
87.     string targetDirectory = Console.ReadLine();
    
88. 
    - u3 K, B, i* T4 n! s
89.     if (!Directory.Exists(sourceDirectory) || !Directory.Exists(targetDirectory))
    
90.     {
      X4 I2 _) Q( F9 w$ m8 u7 [2 _( q
91.         Console.WriteLine("目录不存在，请检查输入的目录路径。");
    - ]! \! R9 [3 L$ N( O3 q7 h. [
92.         return;
    4 v# }0 v2 i7 B* u9 r8 r* J
93.     }
    
94. 
    # o- b. X) O% O1 g: a. O; G
95.     await ProcessFilesAsync(processor ,sourceDirectory, targetDirectory);
    4 r/ Z0 J" i0 A# J5 ^; d
96. 
    
97.     Console.WriteLine("处理完成！");
    
98. 
    
99. }
    ' L! B2 m5 K6 G- l' f  ]
100. static async Task ProcessFilesAsync(WhisperProcessor processor,
     3 \! S0 k0 I# C9 R6 ~" E, e3 z
101.     string sourceDirectory, string targetDirectory)
     
102. {
     ( W* P2 d1 L8 i) }# T# q% p
103.     var files = Directory.GetFiles(sourceDirectory, "*.wav", SearchOption.AllDirectories);
     & o& N" {1 ^1 S6 Y
104. 
     0 h6 ]  p$ K0 j1 F" E
105.     foreach (var sourceFilePath in files)
     & V* U0 }8 H2 g
106.     {
     
107.         string relativePath = Path.GetRelativePath(sourceDirectory, sourceFilePath);
     " T1 D' z5 Q# o
108.         string destinationFilePath = Path.Combine(targetDirectory, relativePath);
     & N& R  a; ?% j* Q2 ~( q$ Q
109.         destinationFilePath = Path.ChangeExtension(destinationFilePath, ".srt");
     
110. 
     * H) o: o* Q* y3 [; l' Y) u
111.         Directory.CreateDirectory(Path.GetDirectoryName(destinationFilePath));
     $ j$ o) |% A9 n
112. 
     : Q: j( d0 E' Q; }6 W
113. 
     + y3 Z  y! E& s. N
114.         if (!File.Exists(destinationFilePath))
     
115.         {
     ' c( c% x3 D: ~1 C) j
116.             Console.WriteLine([        DISCUZ_CODE_1        ]quot;正在处理文件：{sourceFilePath}");
     / E4 n0 `1 D1 ?) r
117. 
     
118.             using var fileStream = File.OpenRead(sourceFilePath);
     : z/ E2 A1 x0 L7 u0 y# K
119.             var segmentIndex = 1;
     
120.             using var writer = new StreamWriter(destinationFilePath); // 创建用于写入srt文件的StreamWriter
     
121.             var startTime = DateTime.Now; // 记录开始时间
     4 v6 E! @7 M- k- u
122. 
     , W3 {, T$ |4 U$ H9 r! g  s8 ]
123.             await foreach (var segment in processor.ProcessAsync(fileStream, CancellationToken.None))
     + \; {& {' c4 c& K' \. d/ u
124.             {
     
125.                 Console.WriteLine([        DISCUZ_CODE_1        ]quot;{segmentIndex}");
     8 _# h8 s5 Q; d7 x3 X7 R1 V
126.                 Console.WriteLine([        DISCUZ_CODE_1        ]quot;{segment.Start:hh\\:mm\\:ss\\,fff} --> {segment.End:hh\\:mm\\:ss\\,fff}");
     - c, U8 s. H" a+ B
127.                 Console.WriteLine(segment.Text);
     
128.                 Console.WriteLine();
     $ h  ^. R6 @3 [' `' H
129. 
     % m1 w0 U% t" X8 Y- ?
130.                 // 将srt内容写入文件
     % s8 n5 i+ i3 M4 F, |. A
131.                 await writer.WriteLineAsync([        DISCUZ_CODE_1        ]quot;{segmentIndex}");
     
132.                 await writer.WriteLineAsync([        DISCUZ_CODE_1        ]quot;{segment.Start:hh\\:mm\\:ss\\,fff} --> {segment.End:hh\\:mm\\:ss\\,fff}");
     
133.                 await writer.WriteLineAsync(segment.Text);
     
134.                 await writer.WriteLineAsync();
     $ K& M8 {9 F* c4 s$ j
135. 
     
136.                 writer.Flush(); // 立即保存srt文件
     
137. 
     & ~, C9 I! b  T3 J7 x& g
138.                 segmentIndex++;
     . ~, A- T3 x9 }! T- k
139.             }
     * n6 X1 v$ |; I. I
140. 
     2 L# I2 W" A3 x/ A% Y& ?4 D
141.             var endTime = DateTime.Now; // 记录结束时间
     
142.             var elapsedMinutes = (endTime - startTime).TotalMinutes;
     . k, v. k: E: _! n# Q% C
143.             Console.WriteLine([        DISCUZ_CODE_1        ]quot;已生成srt文件：{destinationFilePath}");
     3 p# a5 d; ~* w0 a
144.             Console.WriteLine([        DISCUZ_CODE_1        ]quot;生成耗费时间：{elapsedMinutes} 分钟");
     # u, Z5 N# B$ ^4 r5 a) |- L
145.         }
     
146.         else {
     ) A3 V; l3 [- S( ~( q  X
147.             Console.WriteLine([        DISCUZ_CODE_1        ]quot;srt文件已经存在：{destinationFilePath}");
     8 C9 P# L# ]  e3 u6 n
148.         }
     
149.     }
     1 a  T6 Z/ r7 ~5 N  @
150. }
     - k& M* l( e, z' n# V' D2 u& b- h
151. public class Options
     
152. {
     
153.     [Option('t', "command", Required = false, HelpText = "Command to run (lang-detect, transcribe or translate)", Default = "transcribe")]
     ( y$ [& _6 }9 l2 u) L, I9 n) C
154.     public string Command { get; set; }
     6 P# i% w5 a8 E/ ~. M* O
155. 
     
156.     [Option('f', "file", Required = false, HelpText = "File to process", Default = "test.mp3")]
     " L* f$ O9 s6 D, d" i
157.     public string FileName { get; set; }
     
158. 
     
159.     [Option('l', "lang", Required = false, HelpText = "Language", Default = "auto")]
     
160.     public string Language { get; set; }
     
161. 
     6 n: v/ n- \" V- L9 c4 ~6 f4 \
162.     [Option('m', "modelFile", Required = false, HelpText = "Model to use (filename", Default = "ggml-large.bin")]
     
163.     public string ModelName { get; set; }
     
164. 
     
165.     [Option('g', "ggml", Required = false, HelpText = "Ggml Model type to download (if not exists)", Default = GgmlType.Base)]
     
166.     public GgmlType ModelType { get; set; }
     
167. }
     

复制代码

9 Z$ H# R( w4 _4 }/ H( [
) x# [- U# m- s0 t1 t- x