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

Work on tts

Browse Source
This commit is contained in:
Nikolaj Olsson 2024-04-13 07:36:17 +02:00
parent c0ed5acc33
commit 3fec01a167
1 changed files with 127 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

152
src/ui/Forms/Tts/TextToSpeech.cs
View File

@ -21,6 +21,7 @@ namespace Nikse.SubtitleEdit.Forms.Tts
private string _waveFolder; private string _waveFolder;
private readonly List<ActorAndVoice> _actorAndVoices; private readonly List<ActorAndVoice> _actorAndVoices;
private readonly SubtitleFormat _subtitleFormat; private readonly SubtitleFormat _subtitleFormat;
private bool _abort;
public class ActorAndVoice public class ActorAndVoice
{ {
@ -92,14 +93,39 @@ namespace Nikse.SubtitleEdit.Forms.Tts
private void ButtonGenerateTtsClick(object sender, EventArgs e) private void ButtonGenerateTtsClick(object sender, EventArgs e)
{ {
if (buttonGenerateTTS.Text == LanguageSettings.Current.General.Cancel)
{
buttonGenerateTTS.Enabled = false;
_abort = true;
Application.DoEvents();
return;
}
buttonGenerateTTS.Text = LanguageSettings.Current.General.Cancel;
_waveFolder = Path.Combine(Path.GetTempPath(), Guid.NewGuid().ToString()); _waveFolder = Path.Combine(Path.GetTempPath(), Guid.NewGuid().ToString());
Directory.CreateDirectory(_waveFolder); Directory.CreateDirectory(_waveFolder);
GenerateParagraphAudio(_subtitle, true); GenerateParagraphAudio(_subtitle, true, null);
if (_abort)
{
HandleAbort();
return;
}
var fileNames = FixParagraphAudioSpeed(); var fileNames = FixParagraphAudioSpeed();
if (_abort)
{
HandleAbort();
return;
}
var tempAudioFile = MergeAudioParagraphs(fileNames); var tempAudioFile = MergeAudioParagraphs(fileNames);
if (_abort)
{
HandleAbort();
return;
}
// rename result file // rename result file
var resultAudioFileName = Path.Combine(Path.GetDirectoryName(tempAudioFile), Path.GetFileNameWithoutExtension(_videoFileName) + ".wav"); var resultAudioFileName = Path.Combine(Path.GetDirectoryName(tempAudioFile), Path.GetFileNameWithoutExtension(_videoFileName) + ".wav");
@ -110,28 +136,49 @@ namespace Nikse.SubtitleEdit.Forms.Tts
if (checkBoxAddToVideoFile.Checked) if (checkBoxAddToVideoFile.Checked)
{ {
AddAudioToVideoFile(resultAudioFileName); AddAudioToVideoFile(resultAudioFileName);
if (_abort)
{
HandleAbort();
return;
}
} }
HandleAbort();
UiUtil.OpenFolder(_waveFolder); UiUtil.OpenFolder(_waveFolder);
} }
private void GenerateParagraphAudio(Subtitle subtitle, bool showProgressBar) private void HandleAbort()
{
buttonGenerateTTS.Enabled = false;
progressBar1.Visible = false;
labelProgress.Text = string.Empty;
if (_abort)
{
Cleanup(_waveFolder, string.Empty);
}
buttonGenerateTTS.Text = "Generate text to speech";
buttonGenerateTTS.Enabled = true;
_abort = false;
}
private void GenerateParagraphAudio(Subtitle subtitle, bool showProgressBar, string overrideFileName)
{ {
if (nikseComboBoxEngine.SelectedIndex == 0) if (nikseComboBoxEngine.SelectedIndex == 0)
{ {
GenerateParagraphAudioMs(subtitle, showProgressBar); GenerateParagraphAudioMs(subtitle, showProgressBar, overrideFileName);
} }
else if (nikseComboBoxEngine.SelectedIndex == 1) else if (nikseComboBoxEngine.SelectedIndex == 1)
{ {
GenerateParagraphAudioPiperTts(subtitle, showProgressBar); GenerateParagraphAudioPiperTts(subtitle, showProgressBar, overrideFileName);
} }
else if (nikseComboBoxEngine.SelectedIndex == 2) else if (nikseComboBoxEngine.SelectedIndex == 2)
{ {
GenerateParagraphAudioTortoiseTts(subtitle, showProgressBar); GenerateParagraphAudioTortoiseTts(subtitle, showProgressBar, overrideFileName);
} }
else if (nikseComboBoxEngine.SelectedIndex == 3) else if (nikseComboBoxEngine.SelectedIndex == 3)
{ {
GenerateParagraphAudioMimic3(subtitle, showProgressBar); GenerateParagraphAudioMimic3(subtitle, showProgressBar, overrideFileName);
} }
} }
@ -147,7 +194,14 @@ namespace Nikse.SubtitleEdit.Forms.Tts
var outputFileName = Path.Combine(_waveFolder, Path.GetFileNameWithoutExtension(audioFileName) + videoExt); var outputFileName = Path.Combine(_waveFolder, Path.GetFileNameWithoutExtension(audioFileName) + videoExt);
var addAudioProcess = VideoPreviewGenerator.AddAudioTrack(_videoFileName, audioFileName, outputFileName); var addAudioProcess = VideoPreviewGenerator.AddAudioTrack(_videoFileName, audioFileName, outputFileName);
addAudioProcess.Start(); addAudioProcess.Start();
addAudioProcess.WaitForExit(); while (!addAudioProcess.HasExited)
{
Application.DoEvents();
if (_abort)
{
break;
}
}
labelProgress.Text = string.Empty; labelProgress.Text = string.Empty;
} }
@ -186,7 +240,13 @@ namespace Nikse.SubtitleEdit.Forms.Tts
var outputFileName1 = Path.Combine(_waveFolder, index + "_u.wav"); var outputFileName1 = Path.Combine(_waveFolder, index + "_u.wav");
var trimProcess = VideoPreviewGenerator.TrimSilenceStartAndEnd(pFileName, outputFileName1); var trimProcess = VideoPreviewGenerator.TrimSilenceStartAndEnd(pFileName, outputFileName1);
trimProcess.Start(); trimProcess.Start();
trimProcess.WaitForExit(); while (!trimProcess.HasExited)
{
if (_abort)
{
return new List<string>();
}
}
var addDuration = 0d; var addDuration = 0d;
if (next != null && p.EndTime.TotalMilliseconds < next.StartTime.TotalMilliseconds) if (next != null && p.EndTime.TotalMilliseconds < next.StartTime.TotalMilliseconds)
@ -199,6 +259,11 @@ namespace Nikse.SubtitleEdit.Forms.Tts
} }
} }
if (_abort)
{
return new List<string>();
}
var waveInfo = UiUtil.GetVideoInfo(outputFileName1); var waveInfo = UiUtil.GetVideoInfo(outputFileName1);
if (waveInfo.TotalMilliseconds <= p.DurationTotalMilliseconds + addDuration) if (waveInfo.TotalMilliseconds <= p.DurationTotalMilliseconds + addDuration)
{ {
@ -211,7 +276,15 @@ namespace Nikse.SubtitleEdit.Forms.Tts
fileNames.Add(outputFileName2); fileNames.Add(outputFileName2);
var mergeProcess = VideoPreviewGenerator.ChangeSpeed(outputFileName1, outputFileName2, (float)factor); var mergeProcess = VideoPreviewGenerator.ChangeSpeed(outputFileName1, outputFileName2, (float)factor);
mergeProcess.Start(); mergeProcess.Start();
mergeProcess.WaitForExit();
while (!mergeProcess.HasExited)
{
Application.DoEvents();
if (_abort)
{
return new List<string>();
}
}
} }
return fileNames; return fileNames;
@ -248,7 +321,15 @@ namespace Nikse.SubtitleEdit.Forms.Tts
var mergeProcess = VideoPreviewGenerator.MergeAudioTracks(inputFileName, pFileName, outputFileName, (float)p.StartTime.TotalSeconds); var mergeProcess = VideoPreviewGenerator.MergeAudioTracks(inputFileName, pFileName, outputFileName, (float)p.StartTime.TotalSeconds);
inputFileName = outputFileName; inputFileName = outputFileName;
mergeProcess.Start(); mergeProcess.Start();
mergeProcess.WaitForExit();
while (!mergeProcess.HasExited)
{
Application.DoEvents();
if (_abort)
{
return string.Empty;
}
}
} }
progressBar1.Visible = false; progressBar1.Visible = false;
@ -257,7 +338,7 @@ namespace Nikse.SubtitleEdit.Forms.Tts
return outputFileName; return outputFileName;
} }
private void GenerateParagraphAudioMs(Subtitle subtitle, bool showProgressBar) private void GenerateParagraphAudioMs(Subtitle subtitle, bool showProgressBar, string overrideFileName)
{ {
progressBar1.Value = 0; progressBar1.Value = 0;
progressBar1.Maximum = subtitle.Paragraphs.Count; progressBar1.Maximum = subtitle.Paragraphs.Count;
@ -286,7 +367,8 @@ namespace Nikse.SubtitleEdit.Forms.Tts
} }
var p = subtitle.Paragraphs[index]; var p = subtitle.Paragraphs[index];
synthesizer.SetOutputToWaveFile(Path.Combine(_waveFolder, index + ".wav")); var wavFileName = Path.Combine(_waveFolder, string.IsNullOrEmpty(overrideFileName) ? index + ".wav" : overrideFileName);
synthesizer.SetOutputToWaveFile(wavFileName);
var builder = new PromptBuilder(); var builder = new PromptBuilder();
if (voiceInfo != null) if (voiceInfo != null)
{ {
@ -322,7 +404,7 @@ namespace Nikse.SubtitleEdit.Forms.Tts
labelProgress.Text = string.Empty; labelProgress.Text = string.Empty;
} }
private bool GenerateParagraphAudioPiperTts(Subtitle subtitle, bool showProgressBar) private bool GenerateParagraphAudioPiperTts(Subtitle subtitle, bool showProgressBar, string overrideFileName)
{ {
var ttsPath = Path.Combine(Configuration.DataDirectory, "TextToSpeech"); var ttsPath = Path.Combine(Configuration.DataDirectory, "TextToSpeech");
if (!Directory.Exists(ttsPath)) if (!Directory.Exists(ttsPath))
@ -338,7 +420,7 @@ namespace Nikse.SubtitleEdit.Forms.Tts
var piperExe = Path.Combine(piperPath, "piper.exe"); var piperExe = Path.Combine(piperPath, "piper.exe");
if (!File.Exists(piperExe)) if (Configuration.IsRunningOnWindows && !File.Exists(piperExe))
{ {
if (MessageBox.Show(string.Format(LanguageSettings.Current.Settings.DownloadX, "Piper Text To Speech"), "Subtitle Edit", MessageBoxButtons.YesNoCancel) != DialogResult.Yes) if (MessageBox.Show(string.Format(LanguageSettings.Current.Settings.DownloadX, "Piper Text To Speech"), "Subtitle Edit", MessageBoxButtons.YesNoCancel) != DialogResult.Yes)
{ {
@ -368,14 +450,14 @@ namespace Nikse.SubtitleEdit.Forms.Tts
} }
var p = subtitle.Paragraphs[index]; var p = subtitle.Paragraphs[index];
var outputFileName = Path.Combine(_waveFolder, index + ".wav"); var outputFileName = Path.Combine(_waveFolder, string.IsNullOrEmpty(overrideFileName) ? index + ".wav" : overrideFileName);
if (File.Exists(outputFileName)) if (File.Exists(outputFileName))
{ {
try try
{ {
File.Delete(outputFileName); File.Delete(outputFileName);
} }
catch catch
{ {
// ignore // ignore
} }
@ -420,7 +502,7 @@ namespace Nikse.SubtitleEdit.Forms.Tts
StartInfo = StartInfo =
{ {
WorkingDirectory = piperPath, WorkingDirectory = piperPath,
FileName = piperExe, FileName = Configuration.IsRunningOnWindows ? piperExe : "piper",
Arguments = $"-m \"{voice.ModelShort}\" -c \"{voice.ConfigShort}\" -f out.wav", Arguments = $"-m \"{voice.ModelShort}\" -c \"{voice.ConfigShort}\" -f out.wav",
UseShellExecute = false, UseShellExecute = false,
CreateNoWindow = true, CreateNoWindow = true,
@ -433,7 +515,17 @@ namespace Nikse.SubtitleEdit.Forms.Tts
streamWriter.Write(p.Text); streamWriter.Write(p.Text);
streamWriter.Flush(); streamWriter.Flush();
streamWriter.Close(); streamWriter.Close();
processPiper.WaitForExit();
while (!processPiper.HasExited)
{
Application.DoEvents();
if (_abort)
{
progressBar1.Visible = false;
labelProgress.Text = string.Empty;
return false;
}
}
var inputFile = Path.Combine(piperPath, "out.wav"); var inputFile = Path.Combine(piperPath, "out.wav");
File.Move(inputFile, outputFileName); File.Move(inputFile, outputFileName);
@ -449,7 +541,7 @@ namespace Nikse.SubtitleEdit.Forms.Tts
return true; return true;
} }
private bool GenerateParagraphAudioTortoiseTts(Subtitle subtitle, bool showProgressBar) private bool GenerateParagraphAudioTortoiseTts(Subtitle subtitle, bool showProgressBar, string overrideFileName)
{ {
var pythonFolder = Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.UserProfile), var pythonFolder = Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.UserProfile),
"AppData", "AppData",
@ -486,7 +578,7 @@ namespace Nikse.SubtitleEdit.Forms.Tts
} }
var p = subtitle.Paragraphs[index]; var p = subtitle.Paragraphs[index];
var outputFileName = Path.Combine(_waveFolder, index + ".wav"); var outputFileName = Path.Combine(_waveFolder, string.IsNullOrEmpty(overrideFileName) ? index + ".wav" : overrideFileName);
var v = voice; var v = voice;
if (_actorAndVoices.Count > 0 && !string.IsNullOrEmpty(p.Actor)) if (_actorAndVoices.Count > 0 && !string.IsNullOrEmpty(p.Actor))
@ -511,7 +603,16 @@ namespace Nikse.SubtitleEdit.Forms.Tts
}; };
processTortoiseTts.Start(); processTortoiseTts.Start();
processTortoiseTts.WaitForExit(); while (!processTortoiseTts.HasExited)
{
Application.DoEvents();
if (_abort)
{
progressBar1.Visible = false;
labelProgress.Text = string.Empty;
return false;
}
}
var inputFile = Path.Combine(_waveFolder, $"{v}_0_2.wav"); var inputFile = Path.Combine(_waveFolder, $"{v}_0_2.wav");
File.Move(inputFile, outputFileName); File.Move(inputFile, outputFileName);
@ -527,7 +628,7 @@ namespace Nikse.SubtitleEdit.Forms.Tts
return true; return true;
} }
private void GenerateParagraphAudioMimic3(Subtitle subtitle, bool showProgressBar) private void GenerateParagraphAudioMimic3(Subtitle subtitle, bool showProgressBar, string overrideFileName)
{ {
throw new NotImplementedException(); throw new NotImplementedException();
} }
@ -667,8 +768,9 @@ namespace Nikse.SubtitleEdit.Forms.Tts
var text = TextBoxTest.Text; var text = TextBoxTest.Text;
var sub = new Subtitle(); var sub = new Subtitle();
sub.Paragraphs.Add(new Paragraph(text, 0, 2500)); sub.Paragraphs.Add(new Paragraph(text, 0, 2500));
GenerateParagraphAudio(sub, false); var waveFileNameOnly = Guid.NewGuid() + ".wav";
var waveFileName = Path.Combine(_waveFolder, "0.wav"); GenerateParagraphAudio(sub, false, waveFileNameOnly);
var waveFileName = Path.Combine(_waveFolder, waveFileNameOnly);
using (var soundPlayer = new System.Media.SoundPlayer(waveFileName)) using (var soundPlayer = new System.Media.SoundPlayer(waveFileName))
{ {
soundPlayer.Play(); soundPlayer.Play();
@ -692,7 +794,7 @@ namespace Nikse.SubtitleEdit.Forms.Tts
} }
finally finally
{ {
buttonTestVoice.Enabled = true; TaskDelayHelper.RunDelayed(TimeSpan.FromSeconds(1), () => buttonTestVoice.Enabled = true);
} }
} }
} }