Mirrors/SubtitleEdit
1
0
Fork 0
mirror of https://github.com/SubtitleEdit/subtitleedit.git synced 2024-11-22 03:02:35 +01:00
Code Issues Projects Releases Wiki Activity

Minor fix for whisper UI

Browse Source
This commit is contained in:
niksedk 2022-11-19 13:53:22 +01:00
parent 828cd46efb
commit 7695391c47
12 changed files with 214 additions and 124 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
Changelog.txt
View File

@ -15,6 +15,7 @@
* Update Bulgarian translation - thx Калин
* Update Korean translation - thx domddol
* Update Polish translation - thx admas
* Update Greek translation - thx PMitsakis
* Optimize bd sup saving in batch/cmd-line convert
* Add a few English words/names to dictionaries
* Add more settings for "Delete auto-backup after" - thx pahtashow

16
src/libse/AudioToText/WhisperHelper.cs
View File

@ -8,17 +8,17 @@ namespace Nikse.SubtitleEdit.Core.AudioToText
{
public static IWhisperModel GetWhisperModel()
{
return Configuration.Settings.Tools.UseWhisperCpp ? (IWhisperModel)new WhisperCppModel() : new WhisperModel();
return Configuration.Settings.Tools.WhisperUseCpp ? (IWhisperModel)new WhisperCppModel() : new WhisperModel();
}
public static string ModelExtension()
{
return Configuration.Settings.Tools.UseWhisperCpp ? ".bin" : ".pt";
return Configuration.Settings.Tools.WhisperUseCpp ? ".bin" : ".pt";
}
public static string GetWebSiteUrl()
{
return Configuration.Settings.Tools.UseWhisperCpp ? "https://github.com/ggerganov/whisper.cpp" : "https://github.com/openai/whisper";
return Configuration.Settings.Tools.WhisperUseCpp ? "https://github.com/ggerganov/whisper.cpp" : "https://github.com/openai/whisper";
}
public static bool IsWhisperInstalled()
@ -33,7 +33,7 @@ namespace Nikse.SubtitleEdit.Core.AudioToText
public static string GetWhisperFolder()
{
if (Configuration.IsRunningOnLinux && Configuration.Settings.Tools.UseWhisperCpp)
if (Configuration.IsRunningOnLinux && Configuration.Settings.Tools.WhisperUseCpp)
{
var path = Path.Combine(Configuration.DataDirectory, "Whisper");
return Directory.Exists(path) ? path : null;
@ -60,7 +60,7 @@ namespace Nikse.SubtitleEdit.Core.AudioToText
}
}
if (Configuration.Settings.Tools.UseWhisperCpp)
if (Configuration.Settings.Tools.WhisperUseCpp)
{
var path = Path.Combine(Configuration.DataDirectory, "Whisper");
return Directory.Exists(path) ? path : null;
@ -100,7 +100,7 @@ namespace Nikse.SubtitleEdit.Core.AudioToText
if (Configuration.IsRunningOnWindows)
{
if (Configuration.Settings.Tools.UseWhisperCpp)
if (Configuration.Settings.Tools.WhisperUseCpp)
{
var f = Path.Combine(GetWhisperFolder(), "main.exe");
if (File.Exists(f))
@ -118,7 +118,7 @@ namespace Nikse.SubtitleEdit.Core.AudioToText
}
}
if (Configuration.IsRunningOnLinux && Configuration.Settings.Tools.UseWhisperCpp)
if (Configuration.IsRunningOnLinux && Configuration.Settings.Tools.WhisperUseCpp)
{
var f = Path.Combine(GetWhisperFolder(), "main");
if (File.Exists(f))
@ -132,7 +132,7 @@ namespace Nikse.SubtitleEdit.Core.AudioToText
public static string GetWhisperModelForCmdLine(string model)
{
if (Configuration.Settings.Tools.UseWhisperCpp)
if (Configuration.Settings.Tools.WhisperUseCpp)
{
return Path.Combine(GetWhisperModel().ModelFolder, model + ModelExtension());
}

19
src/libse/Common/Settings.cs
View File

@ -410,7 +410,8 @@ namespace Nikse.SubtitleEdit.Core.Common
public bool GenVideoNonAssaFixRtlUnicode { get; set; }
public bool VoskPostProcessing { get; set; }
public string VoskModel { get; set; }
public bool UseWhisperCpp { get; set; }
public bool WhisperUseCpp { get; set; }
public bool WhisperDeleteTempFiles { get; set; }
public string WhisperModel { get; set; }
public string WhisperLanguageCode { get; set; }
public string WhisperLocation { get; set; }
@ -614,7 +615,8 @@ namespace Nikse.SubtitleEdit.Core.Common
GenVideoFontSizePercentOfHeight = 0.078f;
GenVideoNonAssaBox = true;
VoskPostProcessing = true;
UseWhisperCpp = Configuration.IsRunningOnWindows;
WhisperUseCpp = Configuration.IsRunningOnWindows;
WhisperDeleteTempFiles = true;
WhisperExtraSettings = "";
WhisperLanguageCode = "en";
AudioToTextLineMaxChars = 86;
@ -6131,10 +6133,16 @@ $HorzAlign = Center
settings.Tools.VoskModel = subNode.InnerText;
}
subNode = node.SelectSingleNode("UseWhisperCpp");
subNode = node.SelectSingleNode("WhisperUseCpp");
if (subNode != null)
{
settings.Tools.UseWhisperCpp = Convert.ToBoolean(subNode.InnerText, CultureInfo.InvariantCulture);
settings.Tools.WhisperUseCpp = Convert.ToBoolean(subNode.InnerText, CultureInfo.InvariantCulture);
}
subNode = node.SelectSingleNode("WhisperDeleteTempFiles");
if (subNode != null)
{
settings.Tools.WhisperDeleteTempFiles = Convert.ToBoolean(subNode.InnerText, CultureInfo.InvariantCulture);
}
subNode = node.SelectSingleNode("WhisperModel");
@ -10406,7 +10414,8 @@ $HorzAlign = Center
textWriter.WriteElementString("GenVideoNonAssaFixRtlUnicode", settings.Tools.GenVideoNonAssaFixRtlUnicode.ToString(CultureInfo.InvariantCulture));
textWriter.WriteElementString("VoskPostProcessing", settings.Tools.VoskPostProcessing.ToString(CultureInfo.InvariantCulture));
textWriter.WriteElementString("VoskModel", settings.Tools.VoskModel);
textWriter.WriteElementString("UseWhisperCpp", settings.Tools.UseWhisperCpp.ToString(CultureInfo.InvariantCulture));
textWriter.WriteElementString("WhisperUseCpp", settings.Tools.WhisperUseCpp.ToString(CultureInfo.InvariantCulture));
textWriter.WriteElementString("WhisperDeleteTempFiles", settings.Tools.WhisperDeleteTempFiles.ToString(CultureInfo.InvariantCulture));
textWriter.WriteElementString("WhisperModel", settings.Tools.WhisperModel);
textWriter.WriteElementString("WhisperLocation", settings.Tools.WhisperLocation);
textWriter.WriteElementString("WhisperExtraSettings", settings.Tools.WhisperExtraSettings);

2
src/ui/Forms/AudioToText/VoskAudioToText.cs
View File

@ -143,7 +143,6 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
buttonBatchMode.Enabled = false;
comboBoxModels.Enabled = false;
var waveFileName = GenerateWavFile(_videoFileName, _audioTrackNumber);
textBoxLog.AppendText("Wav file name: " + waveFileName);
textBoxLog.AppendText(Environment.NewLine);
progressBar1.Style = ProgressBarStyle.Blocks;
var transcript = TranscribeViaVosk(waveFileName, modelFileName);
@ -201,7 +200,6 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
continue;
}
textBoxLog.AppendText("Wav file name: " + waveFileName + Environment.NewLine);
progressBar1.Style = ProgressBarStyle.Blocks;
var transcript = TranscribeViaVosk(waveFileName, modelFileName);
if (_cancel)

1
src/ui/Forms/AudioToText/VoskAudioToTextSelectedLines.cs
View File

@ -114,7 +114,6 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
buttonDownload.Enabled = false;
comboBoxModels.Enabled = false;
var waveFileName = videoFileName;
textBoxLog.AppendText("Wav file name: " + waveFileName + Environment.NewLine);
var transcript = TranscribeViaVosk(waveFileName, modelFileName);
if (_cancel)
{

52
src/ui/Forms/AudioToText/WhisperAudioToText.Designer.cs generated
View File

@ -58,8 +58,10 @@
this.labelCpp = new System.Windows.Forms.Label();
this.labelElapsed = new System.Windows.Forms.Label();
this.contextMenuStripWhisperAdvanced = new System.Windows.Forms.ContextMenuStrip(this.components);
this.speedupToolStripMenuItem = new System.Windows.Forms.ToolStripMenuItem();
this.noneToolStripMenuItem = new System.Windows.Forms.ToolStripMenuItem();
this.whisperPhpOriginalToolStripMenuItem = new System.Windows.Forms.ToolStripMenuItem();
this.whisperCppCToolStripMenuItem = new System.Windows.Forms.ToolStripMenuItem();
this.toolStripSeparator1 = new System.Windows.Forms.ToolStripSeparator();
this.removeTemporaryFilesToolStripMenuItem = new System.Windows.Forms.ToolStripMenuItem();
this.groupBoxModels.SuspendLayout();
this.groupBoxInputFiles.SuspendLayout();
this.contextMenuStripWhisperAdvanced.SuspendLayout();
@ -373,24 +375,38 @@
// contextMenuStripWhisperAdvanced
//
this.contextMenuStripWhisperAdvanced.Items.AddRange(new System.Windows.Forms.ToolStripItem[] {
this.noneToolStripMenuItem,
this.speedupToolStripMenuItem});
this.whisperPhpOriginalToolStripMenuItem,
this.whisperCppCToolStripMenuItem,
this.toolStripSeparator1,
this.removeTemporaryFilesToolStripMenuItem});
this.contextMenuStripWhisperAdvanced.Name = "contextMenuStripWhisperAdvanced";
this.contextMenuStripWhisperAdvanced.Size = new System.Drawing.Size(181, 70);
this.contextMenuStripWhisperAdvanced.Size = new System.Drawing.Size(200, 98);
//
// speedupToolStripMenuItem
// whisperPhpOriginalToolStripMenuItem
//
this.speedupToolStripMenuItem.Name = "speedupToolStripMenuItem";
this.speedupToolStripMenuItem.Size = new System.Drawing.Size(180, 22);
this.speedupToolStripMenuItem.Text = "--speed-up";
this.speedupToolStripMenuItem.Click += new System.EventHandler(this.speedupToolStripMenuItem_Click);
this.whisperPhpOriginalToolStripMenuItem.Name = "whisperPhpOriginalToolStripMenuItem";
this.whisperPhpOriginalToolStripMenuItem.Size = new System.Drawing.Size(199, 22);
this.whisperPhpOriginalToolStripMenuItem.Text = "Whisper php (original)";
this.whisperPhpOriginalToolStripMenuItem.Click += new System.EventHandler(this.whisperPhpOriginalToolStripMenuItem_Click);
//
// noneToolStripMenuItem
// whisperCppCToolStripMenuItem
//
this.noneToolStripMenuItem.Name = "noneToolStripMenuItem";
this.noneToolStripMenuItem.Size = new System.Drawing.Size(180, 22);
this.noneToolStripMenuItem.Text = "(none)";
this.noneToolStripMenuItem.Click += new System.EventHandler(this.noneToolStripMenuItem_Click);
this.whisperCppCToolStripMenuItem.Name = "whisperCppCToolStripMenuItem";
this.whisperCppCToolStripMenuItem.Size = new System.Drawing.Size(199, 22);
this.whisperCppCToolStripMenuItem.Text = "Whisper cpp (C++)";
this.whisperCppCToolStripMenuItem.Click += new System.EventHandler(this.whisperCppCToolStripMenuItem_Click);
//
// toolStripSeparator1
//
this.toolStripSeparator1.Name = "toolStripSeparator1";
this.toolStripSeparator1.Size = new System.Drawing.Size(196, 6);
//
// removeTemporaryFilesToolStripMenuItem
//
this.removeTemporaryFilesToolStripMenuItem.Name = "removeTemporaryFilesToolStripMenuItem";
this.removeTemporaryFilesToolStripMenuItem.Size = new System.Drawing.Size(199, 22);
this.removeTemporaryFilesToolStripMenuItem.Text = "Remove temporary files";
this.removeTemporaryFilesToolStripMenuItem.Click += new System.EventHandler(this.removeTemporaryFilesToolStripMenuItem_Click);
//
// WhisperAudioToText
//
@ -465,7 +481,9 @@
private System.Windows.Forms.Label labelCpp;
private System.Windows.Forms.Label labelElapsed;
private System.Windows.Forms.ContextMenuStrip contextMenuStripWhisperAdvanced;
private System.Windows.Forms.ToolStripMenuItem speedupToolStripMenuItem;
private System.Windows.Forms.ToolStripMenuItem noneToolStripMenuItem;
private System.Windows.Forms.ToolStripMenuItem whisperPhpOriginalToolStripMenuItem;
private System.Windows.Forms.ToolStripMenuItem whisperCppCToolStripMenuItem;
private System.Windows.Forms.ToolStripSeparator toolStripSeparator1;
private System.Windows.Forms.ToolStripMenuItem removeTemporaryFilesToolStripMenuItem;
}
}

109
src/ui/Forms/AudioToText/WhisperAudioToText.cs
View File

@ -22,7 +22,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
private bool _cancel;
private bool _batchMode;
private int _batchFileNumber;
private readonly List<string> _filesToDelete;
private List<string> _filesToDelete;
private readonly Form _parentForm;
private bool _useCenterChannelOnly;
private int _initialWidth = 725;
@ -70,17 +70,6 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
checkBoxUsePostProcessing.Checked = Configuration.Settings.Tools.VoskPostProcessing;
comboBoxLanguages.Items.Clear();
comboBoxLanguages.Items.AddRange(WhisperLanguage.Languages.OrderBy(p => p.Name).ToArray<object>());
var lang = WhisperLanguage.Languages.FirstOrDefault(p => p.Code == Configuration.Settings.Tools.WhisperLanguageCode);
comboBoxLanguages.Text = lang != null ? lang.ToString() : "English";
FillModels(comboBoxModels, string.Empty);
textBoxLog.Visible = false;
textBoxLog.Dock = DockStyle.Fill;
labelProgress.Text = string.Empty;
labelTime.Text = string.Empty;
_filesToDelete = new List<string>();
if (string.IsNullOrEmpty(videoFileName))
@ -93,21 +82,32 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
listViewInputFiles.Items.Add(videoFileName);
}
textBoxLog.Visible = false;
textBoxLog.Dock = DockStyle.Fill;
labelProgress.Text = string.Empty;
labelTime.Text = string.Empty;
listViewInputFiles.Visible = false;
labelFC.Text = string.Empty;
labelCpp.Visible = Configuration.Settings.Tools.UseWhisperCpp;
labelElapsed.Text = string.Empty;
//if (Configuration.Settings.Tools.UseWhisperCpp)
//{
// speedupToolStripMenuItem.Checked = Configuration.Settings.Tools.WhisperExtraSettings == speedupToolStripMenuItem.Text;
// ContextMenuStrip = contextMenuStripWhisperAdvanced;{}
Init();
}
// if (string.IsNullOrEmpty(Configuration.Settings.Tools.WhisperExtraSettings))
// {
// noneToolStripMenuItem_Click(null, null);
// }
//}
private void Init()
{
comboBoxLanguages.Items.Clear();
comboBoxLanguages.Items.AddRange(WhisperLanguage.Languages.OrderBy(p => p.Name).ToArray<object>());
var lang = WhisperLanguage.Languages.FirstOrDefault(p => p.Code == Configuration.Settings.Tools.WhisperLanguageCode);
comboBoxLanguages.Text = lang != null ? lang.ToString() : "English";
FillModels(comboBoxModels, string.Empty);
labelFC.Text = string.Empty;
labelCpp.Visible = Configuration.Settings.Tools.WhisperUseCpp;
whisperCppCToolStripMenuItem.Checked = Configuration.Settings.Tools.WhisperUseCpp;
whisperPhpOriginalToolStripMenuItem.Checked = !Configuration.Settings.Tools.WhisperUseCpp;
removeTemporaryFilesToolStripMenuItem.Checked = Configuration.Settings.Tools.WhisperDeleteTempFiles;
ContextMenuStrip = contextMenuStripWhisperAdvanced;
}
public static void FillModels(ComboBox comboBoxModels, string lastDownloadedModel)
@ -179,7 +179,6 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
comboBoxLanguages.Enabled = false;
comboBoxModels.Enabled = false;
var waveFileName = GenerateWavFile(_videoFileName, _audioTrackNumber);
_outputText.Add("Wav file name: " + waveFileName);
progressBar1.Style = ProgressBarStyle.Blocks;
timer1.Start();
var transcript = TranscribeViaWhisper(waveFileName);
@ -251,7 +250,6 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
}
_outputText.Add(string.Empty);
_outputText.Add("Wav file name: " + waveFileName);
progressBar1.Style = ProgressBarStyle.Blocks;
var transcript = TranscribeViaWhisper(waveFileName);
if (_cancel)
@ -339,7 +337,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
_resultList = new List<ResultText>();
var process = GetWhisperProcess(waveFileName, model.Name, _languageCode, checkBoxTranslateToEnglish.Checked, OutputHandler);
var sw = Stopwatch.StartNew();
_outputText.Add($"Calling whisper{(Configuration.Settings.Tools.UseWhisperCpp ? "-CPP" : string.Empty)} with : {process.StartInfo.FileName} {process.StartInfo.Arguments}");
_outputText.Add($"Calling whisper{(Configuration.Settings.Tools.WhisperUseCpp ? "-CPP" : string.Empty)} with : {process.StartInfo.FileName} {process.StartInfo.Arguments}");
_startTicks = DateTime.UtcNow.Ticks;
_videoInfo = UiUtil.GetVideoInfo(waveFileName);
timer1.Start();
@ -378,7 +376,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
}
}
_outputText.Add($"Calling whisper{(Configuration.Settings.Tools.UseWhisperCpp ? "-CPP" : string.Empty)} done in {sw.Elapsed}{Environment.NewLine}");
_outputText.Add($"Calling whisper{(Configuration.Settings.Tools.WhisperUseCpp ? "-CPP" : string.Empty)} done in {sw.Elapsed}{Environment.NewLine}");
for (var i = 0; i < 10; i++)
{
@ -386,7 +384,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
System.Threading.Thread.Sleep(50);
}
if (GetResultFromSrt(waveFileName, out var resultTexts, _outputText))
if (GetResultFromSrt(waveFileName, out var resultTexts, _outputText, _filesToDelete))
{
return resultTexts;
}
@ -395,7 +393,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
return _resultList;
}
public static bool GetResultFromSrt(string waveFileName, out List<ResultText> resultTexts, ConcurrentBag<string> outputText)
public static bool GetResultFromSrt(string waveFileName, out List<ResultText> resultTexts, ConcurrentBag<string> outputText, List<string> filesToDelete)
{
var srtFileName = waveFileName + ".srt";
var vttFileName = Path.Combine(WhisperHelper.GetWhisperFolder(), Path.GetFileName(waveFileName) + ".vtt");
@ -443,21 +441,14 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
resultTexts = results;
try
if (File.Exists(srtFileName))
{
if (!File.Exists(srtFileName))
{
File.Delete(srtFileName);
}
if (!File.Exists(vttFileName))
{
File.Delete(vttFileName);
}
filesToDelete?.Add(srtFileName);
}
catch
if (File.Exists(vttFileName))
{
// ignore
filesToDelete?.Add(vttFileName);
}
return true;
@ -664,7 +655,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
translateToEnglish = string.Empty;
}
if (Configuration.Settings.Tools.UseWhisperCpp && string.IsNullOrEmpty(translateToEnglish) && !Configuration.Settings.Tools.WhisperExtraSettings.Contains("--max-len"))
if (Configuration.Settings.Tools.WhisperUseCpp && string.IsNullOrEmpty(translateToEnglish) && !Configuration.Settings.Tools.WhisperExtraSettings.Contains("--max-len"))
{
var maxChars = (int)Math.Round(Configuration.Settings.General.SubtitleLineMaximumLength * 1.8, MidpointRounding.AwayFromZero);
if (language == "jp")
@ -680,7 +671,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
}
var outputSrt = string.Empty;
if (Configuration.Settings.Tools.UseWhisperCpp)
if (Configuration.Settings.Tools.WhisperUseCpp)
{
outputSrt = "--output-srt ";
@ -751,7 +742,17 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
Configuration.Settings.Tools.VoskPostProcessing = checkBoxUsePostProcessing.Checked;
foreach (var fileName in _filesToDelete)
DeleteTemporaryFiles(_filesToDelete);
}
public static void DeleteTemporaryFiles(List<string> filesToDelete)
{
if (!Configuration.Settings.Tools.WhisperDeleteTempFiles)
{
return;
}
foreach (var fileName in filesToDelete)
{
try
{
@ -1036,18 +1037,22 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
checkBoxTranslateToEnglish.Enabled = comboBoxLanguages.Text.ToLowerInvariant() != "english";
}
private void speedupToolStripMenuItem_Click(object sender, EventArgs e)
private void whisperPhpOriginalToolStripMenuItem_Click(object sender, EventArgs e)
{
noneToolStripMenuItem.Checked = false;
speedupToolStripMenuItem.Checked = true;
Configuration.Settings.Tools.WhisperExtraSettings = speedupToolStripMenuItem.Text;
Configuration.Settings.Tools.WhisperUseCpp = false;
Init();
}
private void noneToolStripMenuItem_Click(object sender, EventArgs e)
private void whisperCppCToolStripMenuItem_Click(object sender, EventArgs e)
{
noneToolStripMenuItem.Checked = true;
speedupToolStripMenuItem.Checked = false;
Configuration.Settings.Tools.WhisperExtraSettings = string.Empty;
Configuration.Settings.Tools.WhisperUseCpp = true;
Init();
}
private void removeTemporaryFilesToolStripMenuItem_Click(object sender, EventArgs e)
{
Configuration.Settings.Tools.WhisperDeleteTempFiles = !Configuration.Settings.Tools.WhisperDeleteTempFiles;
removeTemporaryFilesToolStripMenuItem.Checked = Configuration.Settings.Tools.WhisperDeleteTempFiles;
}
}
}

48
src/ui/Forms/AudioToText/WhisperAudioToTextSelectedLines.Designer.cs generated
View File

@ -51,8 +51,14 @@
this.columnHeaderFileName = ((System.Windows.Forms.ColumnHeader)(new System.Windows.Forms.ColumnHeader()));
this.checkBoxTranslateToEnglish = new System.Windows.Forms.CheckBox();
this.labelCpp = new System.Windows.Forms.Label();
this.contextMenuStripWhisperAdvanced = new System.Windows.Forms.ContextMenuStrip(this.components);
this.whisperPhpOriginalToolStripMenuItem = new System.Windows.Forms.ToolStripMenuItem();
this.whisperCppCToolStripMenuItem = new System.Windows.Forms.ToolStripMenuItem();
this.toolStripSeparator1 = new System.Windows.Forms.ToolStripSeparator();
this.removeTemporaryFilesToolStripMenuItem = new System.Windows.Forms.ToolStripMenuItem();
this.groupBoxModels.SuspendLayout();
this.groupBoxInputFiles.SuspendLayout();
this.contextMenuStripWhisperAdvanced.SuspendLayout();
this.SuspendLayout();
//
// buttonCancel
@ -289,6 +295,42 @@
this.labelCpp.TabIndex = 22;
this.labelCpp.Text = "CPP";
//
// contextMenuStripWhisperAdvanced
//
this.contextMenuStripWhisperAdvanced.Items.AddRange(new System.Windows.Forms.ToolStripItem[] {
this.whisperPhpOriginalToolStripMenuItem,
this.whisperCppCToolStripMenuItem,
this.toolStripSeparator1,
this.removeTemporaryFilesToolStripMenuItem});
this.contextMenuStripWhisperAdvanced.Name = "contextMenuStripWhisperAdvanced";
this.contextMenuStripWhisperAdvanced.Size = new System.Drawing.Size(200, 98);
//
// whisperPhpOriginalToolStripMenuItem
//
this.whisperPhpOriginalToolStripMenuItem.Name = "whisperPhpOriginalToolStripMenuItem";
this.whisperPhpOriginalToolStripMenuItem.Size = new System.Drawing.Size(199, 22);
this.whisperPhpOriginalToolStripMenuItem.Text = "Whisper php (original)";
this.whisperPhpOriginalToolStripMenuItem.Click += new System.EventHandler(this.whisperPhpOriginalToolStripMenuItem_Click);
//
// whisperCppCToolStripMenuItem
//
this.whisperCppCToolStripMenuItem.Name = "whisperCppCToolStripMenuItem";
this.whisperCppCToolStripMenuItem.Size = new System.Drawing.Size(199, 22);
this.whisperCppCToolStripMenuItem.Text = "Whisper cpp (C++)";
this.whisperCppCToolStripMenuItem.Click += new System.EventHandler(this.whisperCppCToolStripMenuItem_Click);
//
// toolStripSeparator1
//
this.toolStripSeparator1.Name = "toolStripSeparator1";
this.toolStripSeparator1.Size = new System.Drawing.Size(196, 6);
//
// removeTemporaryFilesToolStripMenuItem
//
this.removeTemporaryFilesToolStripMenuItem.Name = "removeTemporaryFilesToolStripMenuItem";
this.removeTemporaryFilesToolStripMenuItem.Size = new System.Drawing.Size(199, 22);
this.removeTemporaryFilesToolStripMenuItem.Text = "Remove temporary files";
this.removeTemporaryFilesToolStripMenuItem.Click += new System.EventHandler(this.removeTemporaryFilesToolStripMenuItem_Click);
//
// WhisperAudioToTextSelectedLines
//
this.AutoScaleDimensions = new System.Drawing.SizeF(6F, 13F);
@ -322,6 +364,7 @@
this.groupBoxModels.ResumeLayout(false);
this.groupBoxModels.PerformLayout();
this.groupBoxInputFiles.ResumeLayout(false);
this.contextMenuStripWhisperAdvanced.ResumeLayout(false);
this.ResumeLayout(false);
this.PerformLayout();
@ -351,5 +394,10 @@
private System.Windows.Forms.ComboBox comboBoxLanguages;
private System.Windows.Forms.CheckBox checkBoxTranslateToEnglish;
private System.Windows.Forms.Label labelCpp;
private System.Windows.Forms.ContextMenuStrip contextMenuStripWhisperAdvanced;
private System.Windows.Forms.ToolStripMenuItem whisperPhpOriginalToolStripMenuItem;
private System.Windows.Forms.ToolStripMenuItem whisperCppCToolStripMenuItem;
private System.Windows.Forms.ToolStripSeparator toolStripSeparator1;
private System.Windows.Forms.ToolStripMenuItem removeTemporaryFilesToolStripMenuItem;
}
}

77
src/ui/Forms/AudioToText/WhisperAudioToTextSelectedLines.cs
View File

@ -19,6 +19,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
private int _batchFileNumber;
private readonly List<AudioClipsGet.AudioClip> _audioClips;
private readonly Form _parentForm;
private readonly List<string> _filesToDelete;
private readonly Regex _timeRegex = new Regex(@"^\[\d\d:\d\d[\.,]\d\d\d --> \d\d:\d\d[\.,]\d\d\d\]", RegexOptions.Compiled);
private List<ResultText> _resultList;
private string _languageCode;
@ -33,6 +34,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
UiUtil.FixFonts(this);
UiUtil.FixLargeFonts(this, buttonGenerate);
_parentForm = parentForm;
_filesToDelete = new List<string>();
Text = LanguageSettings.Current.AudioToText.Title;
labelInfo.Text = LanguageSettings.Current.AudioToText.WhisperInfo;
@ -51,12 +53,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
checkBoxUsePostProcessing.Checked = Configuration.Settings.Tools.VoskPostProcessing;
comboBoxLanguages.Items.Clear();
comboBoxLanguages.Items.AddRange(WhisperLanguage.Languages.OrderBy(p => p.Name).ToArray<object>());
var lang = WhisperLanguage.Languages.FirstOrDefault(p => p.Code == Configuration.Settings.Tools.WhisperLanguageCode);
comboBoxLanguages.Text = lang != null ? lang.ToString() : "English";
WhisperAudioToText.FillModels(comboBoxModels, string.Empty);
Init();
textBoxLog.Visible = false;
textBoxLog.Dock = DockStyle.Fill;
@ -65,13 +62,27 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
listViewInputFiles.Visible = true;
_audioClips = audioClips;
progressBar1.Maximum = 100;
labelCpp.Visible = Configuration.Settings.Tools.UseWhisperCpp;
labelCpp.Visible = Configuration.Settings.Tools.WhisperUseCpp;
foreach (var audioClip in audioClips)
{
listViewInputFiles.Items.Add(audioClip.AudioFileName);
}
}
private void Init()
{
comboBoxLanguages.Items.Clear();
comboBoxLanguages.Items.AddRange(WhisperLanguage.Languages.OrderBy(p => p.Name).ToArray<object>());
var lang = WhisperLanguage.Languages.FirstOrDefault(p => p.Code == Configuration.Settings.Tools.WhisperLanguageCode);
comboBoxLanguages.Text = lang != null ? lang.ToString() : "English";
WhisperAudioToText.FillModels(comboBoxModels, string.Empty);
whisperCppCToolStripMenuItem.Checked = Configuration.Settings.Tools.WhisperUseCpp;
whisperPhpOriginalToolStripMenuItem.Checked = !Configuration.Settings.Tools.WhisperUseCpp;
removeTemporaryFilesToolStripMenuItem.Checked = Configuration.Settings.Tools.WhisperDeleteTempFiles;
ContextMenuStrip = contextMenuStripWhisperAdvanced;
}
private void ButtonGenerate_Click(object sender, EventArgs e)
{
if (comboBoxModels.Items.Count == 0)
@ -104,12 +115,12 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
private void GenerateBatch()
{
_languageCode = GetLanguage(comboBoxLanguages.Text);
_languageCode = WhisperAudioToText.GetLanguage(comboBoxLanguages.Text);
groupBoxInputFiles.Enabled = false;
comboBoxLanguages.Enabled = false;
comboBoxModels.Enabled = false;
_batchFileNumber = 0;
var postProcessor = new AudioToTextPostProcessor(GetLanguage(comboBoxModels.Text))
var postProcessor = new AudioToTextPostProcessor(_languageCode)
{
ParagraphMaxChars = Configuration.Settings.General.SubtitleLineMaximumLength * 2,
};
@ -129,7 +140,6 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
var waveFileName = videoFileName;
_outputText.Add(string.Empty);
_outputText.Add("Wav file name: " + waveFileName);
progressBar1.Style = ProgressBarStyle.Blocks;
var transcript = TranscribeViaWhisper(waveFileName);
if (_cancel)
@ -169,7 +179,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
_resultList = new List<ResultText>();
var process = WhisperAudioToText.GetWhisperProcess(waveFileName, model.Name, _languageCode, checkBoxTranslateToEnglish.Checked, OutputHandler);
var sw = Stopwatch.StartNew();
_outputText.Add($"Calling whisper{(Configuration.Settings.Tools.UseWhisperCpp ? "-CPP" : string.Empty)} with : whisper {process.StartInfo.Arguments}");
_outputText.Add($"Calling whisper{(Configuration.Settings.Tools.WhisperUseCpp ? "-CPP" : string.Empty)} with : whisper {process.StartInfo.Arguments}");
ShowProgressBar();
progressBar1.Style = ProgressBarStyle.Marquee;
buttonCancel.Visible = true;
@ -207,7 +217,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
}
}
_outputText.Add($"Calling whisper{(Configuration.Settings.Tools.UseWhisperCpp ? "-CPP" : string.Empty)} done in {sw.Elapsed}{Environment.NewLine}");
_outputText.Add($"Calling whisper{(Configuration.Settings.Tools.WhisperUseCpp ? "-CPP" : string.Empty)} done in {sw.Elapsed}{Environment.NewLine}");
for (var i = 0; i < 10; i++)
{
@ -215,7 +225,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
System.Threading.Thread.Sleep(50);
}
if (WhisperAudioToText.GetResultFromSrt(waveFileName, out var resultTexts, _outputText))
if (WhisperAudioToText.GetResultFromSrt(waveFileName, out var resultTexts, _outputText, null))
{
return resultTexts;
}
@ -297,25 +307,6 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
}
}
internal static string GetLanguage(string text)
{
var languageCodeList = VoskModel.Models.Select(p => p.TwoLetterLanguageCode);
foreach (var languageCode in languageCodeList)
{
if (text.Contains("model-" + languageCode) || text.Contains("model-small-" + languageCode) || text.StartsWith(languageCode, StringComparison.OrdinalIgnoreCase))
{
return languageCode;
}
if (languageCode == "jp" && (text.Contains("model-ja") || text.Contains("model-small-ja")))
{
return languageCode;
}
}
return "en";
}
private void buttonCancel_Click(object sender, EventArgs e)
{
if (buttonGenerate.Enabled)
@ -345,7 +336,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
Configuration.Settings.Tools.WhisperLanguageCode = language.Code;
}
Configuration.Settings.Tools.VoskPostProcessing = checkBoxUsePostProcessing.Checked;
WhisperAudioToText.DeleteTemporaryFiles(_filesToDelete);
}
private void AudioToText_KeyDown(object sender, KeyEventArgs e)
@ -419,7 +410,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
using (var form = new WhisperModelDownload { AutoClose = true })
{
form.ShowDialog(this);
VoskAudioToText.FillModels(comboBoxModels, form.LastDownloadedModel.Name);
WhisperAudioToText.FillModels(comboBoxModels, form.LastDownloadedModel.Name);
}
}
@ -449,5 +440,23 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
{
checkBoxTranslateToEnglish.Enabled = comboBoxLanguages.Text.ToLowerInvariant() != "english";
}
private void whisperPhpOriginalToolStripMenuItem_Click(object sender, EventArgs e)
{
Configuration.Settings.Tools.WhisperUseCpp = false;
Init();
}
private void whisperCppCToolStripMenuItem_Click(object sender, EventArgs e)
{
Configuration.Settings.Tools.WhisperUseCpp = true;
Init();
}
private void removeTemporaryFilesToolStripMenuItem_Click(object sender, EventArgs e)
{
Configuration.Settings.Tools.WhisperDeleteTempFiles = !Configuration.Settings.Tools.WhisperDeleteTempFiles;
removeTemporaryFilesToolStripMenuItem.Checked = Configuration.Settings.Tools.WhisperDeleteTempFiles;
}
}
}

3
src/ui/Forms/AudioToText/WhisperAudioToTextSelectedLines.resx
View File

@ -120,4 +120,7 @@
<metadata name="timer1.TrayLocation" type="System.Drawing.Point, System.Drawing, Version=4.0.0.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a">
<value>17, 17</value>
</metadata>
<metadata name="contextMenuStripWhisperAdvanced.TrayLocation" type="System.Drawing.Point, System.Drawing, Version=4.0.0.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a">
<value>99, 16</value>
</metadata>
</root>

2
src/ui/Forms/AudioToText/WhisperModelDownload.cs
View File

@ -136,7 +136,7 @@ namespace Nikse.SubtitleEdit.Forms.AudioToText
}
buttonDownload.Enabled = true;
labelPleaseWait.Text = string.Format(LanguageSettings.Current.SettingsFfmpeg.XDownloadOk, "Vosk model");
labelPleaseWait.Text = string.Format(LanguageSettings.Current.SettingsFfmpeg.XDownloadOk, "Whisper model");
}
}
}

8
src/ui/Forms/Main.cs
View File

@ -34677,7 +34677,7 @@ namespace Nikse.SubtitleEdit.Forms
return;
}
if (Configuration.Settings.Tools.UseWhisperCpp)
if (Configuration.Settings.Tools.WhisperUseCpp)
{
if (!RequireWhisperCpp())
{
@ -34735,7 +34735,7 @@ namespace Nikse.SubtitleEdit.Forms
private static void CheckWhisperCpp()
{
if (!Configuration.Settings.Tools.UseWhisperCpp)
if (!Configuration.Settings.Tools.WhisperUseCpp)
{
return;
}
@ -34743,13 +34743,13 @@ namespace Nikse.SubtitleEdit.Forms
if (Configuration.IsRunningOnLinux && WhisperHelper.GetWhisperPathAndFileName() == "whisper")
{
SeLogger.Error("UseWhisperCpp changed to 'False' as 'Whisper/whisper' or '/Whisper/main' was not found!");
Configuration.Settings.Tools.UseWhisperCpp = false;
Configuration.Settings.Tools.WhisperUseCpp = false;
}
if (Configuration.IsRunningOnWindows && WhisperHelper.GetWhisperPathAndFileName() == "whisper")
{
SeLogger.Error("UseWhisperCpp changed to 'False' as 'Whisper/whisper.exe' or '/Whisper/main.exe' was not found!");
Configuration.Settings.Tools.UseWhisperCpp = false;
Configuration.Settings.Tools.WhisperUseCpp = false;
}
}