Merge branch 'main' of https://github.com/GewoonJaap/WinWhisper

GewoonJaap · Nov 7, 2023 · ff859dc · ff859dc
2 parents 224155d + 1c9ad10
commit ff859dc
Show file tree

Hide file tree

Showing 6 changed files with 50 additions and 18 deletions.
diff --git a/ConsoleApp/ConsoleApp.csproj b/ConsoleApp/ConsoleApp.csproj
@@ -8,7 +8,7 @@
     <AssemblyName>WinWhisper</AssemblyName>
     <ApplicationIcon>WinWhisper-White_Icon.ico</ApplicationIcon>
     <Platforms>AnyCPU;ARM64</Platforms>
-    <Version>1.2.0</Version>
+    <Version>1.3.0</Version>
     <Description>WinWhisper is a Windows application that uses AI to detect and remove background noise from audio files.</Description>
     <Authors>Jaap</Authors>
     <GeneratePackageOnBuild>True</GeneratePackageOnBuild>
@@ -20,7 +20,7 @@
   </ItemGroup>
 
   <ItemGroup>
-    <PackageReference Include="Clowd.Squirrel" Version="2.9.42" />
+    <PackageReference Include="Clowd.Squirrel" Version="2.10.2" />
     <PackageReference Include="Sentry" Version="3.40.1" />
   </ItemGroup>
 

diff --git a/ConsoleApp/Program.cs b/ConsoleApp/Program.cs
@@ -55,7 +55,7 @@ await videosToConvert.Videos.ForEachAsync(async video =>
                         var audioPath = Extractor.ExtractAudioFromVideoFile(video.VideoPath);
 
                         var audioProcessor = new AudioProcessor();
-                        await audioProcessor.ProcessAudio(audioPath, video.LanguageCode, videosToConvert.SubtitleOutputPath);
+                        await audioProcessor.ProcessAudio(audioPath, video.LanguageCode, videosToConvert.SubtitleOutputPath, video.ShouldTranslate);
                         //remove audioPath file
                         File.Delete(audioPath);
                         Console.WriteLine($"Finished audio processing for video: {video.VideoName}");

diff --git a/Data/VideoToConvertObject.cs b/Data/VideoToConvertObject.cs
@@ -6,11 +6,13 @@ public class VideoToConvertObject
     public string LanguageCode { get; }
 
     public string VideoName { get; }
+    public bool ShouldTranslate { get; }
 
-    public VideoToConvertObject(string videoPath, string languageCode)
+    public VideoToConvertObject(string videoPath, string languageCode, bool shouldTranslate = false)
     {
         VideoPath = videoPath;
         LanguageCode = languageCode;
         VideoName = Path.GetFileName(videoPath);
+        ShouldTranslate = shouldTranslate;
     }
 }
diff --git a/Utility/ConsoleUtil.cs b/Utility/ConsoleUtil.cs
@@ -13,6 +13,13 @@ public static string AskForLanguageCode(string fileName)
         return languageCode;
     }
 
+    public static bool AskIfNeedsToBeTranslated()
+    {
+        Console.WriteLine("Do you want to translate the subtitles to English? (yes/no) Default: no");
+        var translateSubtitles = Console.ReadLine() ?? string.Empty;
+        return translateSubtitles.ToLower() == "yes" || translateSubtitles.ToLower() == "y";
+    }
+
     public static void LogException(Exception ex)
     {
         Console.WriteLine("An error occured. Please report the following on our GitHub page: https://github.com/GewoonJaap/WinWhisper/issues/new?assignees=&labels=&projects=&template=bug_report.md&title=");

diff --git a/Utility/VideoFinder.cs b/Utility/VideoFinder.cs
@@ -52,6 +52,7 @@ private static VideoToConvertObject GetVideoDetails(string videoPath)
     {
         var videoName = Path.GetFileName(videoPath);
         var languageCode = ConsoleUtil.AskForLanguageCode(videoName);
-        return new VideoToConvertObject(videoPath, languageCode);
+        var shouldTranslate = ConsoleUtil.AskIfNeedsToBeTranslated();
+        return new VideoToConvertObject(videoPath, languageCode, shouldTranslate);
     }
 }
diff --git a/WhisperAI/AudioProcessor.cs b/WhisperAI/AudioProcessor.cs
@@ -9,7 +9,7 @@ public class AudioProcessor
         private List<SegmentData> _segments = new();
         private const GgmlType ModelType = GgmlType.Base;
 
-        public async Task ProcessAudio(string wavPath, string languageCode, string subtitleOutputPath)
+        public async Task ProcessAudio(string wavPath, string languageCode, string subtitleOutputPath, bool shouldTranslate)
         {
             var modelName = ModelNameFetcher.GgmlTypeToString(ModelType);
             var modelPath = FolderManager.Models + "/" + modelName;
@@ -24,21 +24,14 @@ public async Task ProcessAudio(string wavPath, string languageCode, string subti
                 fileWriter.Close();
                 Console.WriteLine("Downloaded model");
             }
-
-            var whisperFactory = WhisperFactory.FromPath(modelPath);
 
-            var builder = whisperFactory.CreateBuilder()
-                .WithSegmentEventHandler(OnNewSegment);
+            var processor = SetupProcessor(modelPath, languageCode, shouldTranslate, OnNewSegment);
 
-            if (languageCode.Length > 0)
+            if(processor is null)
             {
-                builder.WithLanguage(languageCode);
+                Console.WriteLine("Something went wrong while setting up the processor. Please try again.");
+                return;
             }
-            else
-            {
-                builder.WithLanguageDetection();
-            }
-            var processor = builder.Build();
 
             void OnNewSegment(SegmentData segmentData)
             {
@@ -54,17 +47,22 @@ void OnNewSegment(SegmentData segmentData)
             {
                 GC.KeepAlive(fileStream);
                 GC.KeepAlive(processor);
+
                 Console.WriteLine("Processing audio file");
+
                 processor.Process(fileStream);
+
                 Console.WriteLine("Processed");
+
                 processor.Dispose();
             }
             //Sort segments by start time
             _segments.Sort((x, y) => x.Start.CompareTo(y.Start));
             //remove all segments that have same time + content
             _segments = _segments.Distinct().ToList();
             //Write segments to file as subtitles
-            var outputFilePath = GetOutputFilePath(wavPath, languageCode.Length == 0 ? "en" : languageCode, subtitleOutputPath);
+            var outputLanguagecode = languageCode.Length == 0 || shouldTranslate ? "en" : languageCode;
+            var outputFilePath = GetOutputFilePath(wavPath, outputLanguagecode, subtitleOutputPath);
 
             var writer = new StreamWriter(outputFilePath);
             var subtitleIndex = 0;
@@ -126,5 +124,29 @@ private static string GetOutputFilePath(string wavPath, string languageCode, str
             var finalOutputPath = outputPath.Length == 0 ? FolderManager.SubtitlesFolder : outputPath;
             return finalOutputPath + "/" + fileName + "." + languageCode.ToUpper() + ".srt";
         }
+
+        private static WhisperProcessor? SetupProcessor(string modelPath, string languageCode, bool shouldTranslate, OnSegmentEventHandler OnNewSegment)
+        {
+            var whisperFactory = WhisperFactory.FromPath(modelPath);
+
+            var builder = whisperFactory.CreateBuilder()
+                .WithSegmentEventHandler(OnNewSegment);
+
+            if (languageCode.Length > 0)
+            {
+                builder.WithLanguage(languageCode);
+            }
+            else
+            {
+                builder.WithLanguageDetection();
+            }
+
+            if (shouldTranslate)
+            {
+                builder.WithTranslate();
+            }
+
+           return builder.Build();
+        }
     }
 }