internal List <Tuple <string, string> > GetAeneasTimings(AutoSegmentRequest requestParameters)
{
string dummy1;
List <string> dummy2;
return(GetAeneasTimings(requestParameters, out dummy1, out dummy2));
}
/// <summary>
/// API Handler when the Auto Segment button is clicked
///
/// Replies with true if AutoSegment completed successfully, or false if there was an error. In addition, a NonFatal message/exception may be reported with the error
/// </summary>
public void AutoSegment(ApiRequest request)
{
Logger.WriteEvent("AudioSegmentationApi.AutoSegment(): AutoSegment started.");
// Parse the JSON containing the text segmentation data.
string json = request.RequiredPostJson();
AutoSegmentRequest requestParameters = ParseJson(json);
string audioFilenameToSegment;
List <string> fragmentIds;
var timingStartEndRangeList = GetAeneasTimings(requestParameters, out audioFilenameToSegment, out fragmentIds);
if (timingStartEndRangeList == null)
{
request.ReplyWithBoolean(false);
return;
}
try
{
ExtractAudioSegments(fragmentIds, timingStartEndRangeList, audioFilenameToSegment);
}
catch (Exception e)
{
ProblemReportApi.ShowProblemDialog(null, e,
"AudioSegmentationApi.AutoSegment(): Exception thrown during split/extract stage", "nonfatal");
request.ReplyWithBoolean(false);
return;
}
Logger.WriteEvent("AudioSegmentationApi.AutoSegment(): Completed successfully.");
request.ReplyWithBoolean(true); // Success
}
/// <summary>
/// "Soft Split": Performs Forced Alignment and responds with the start times of each segment.
/// Output format is a space-separated string of numbers representing the end time (calculated from the beginning of the file) in seconds.
/// </summary>
/// <param name="request"></param>
public void GetForcedAlignmentTimings(ApiRequest request)
{
Logger.WriteEvent("AudioSegmentationApi.GetForcedAlignmentTimings(): GetForcedAlignmentTimings started.");
// Parse the JSON containing the text segmentation data.
string json = request.RequiredPostJson();
AutoSegmentRequest requestParameters = ParseJson(json);
var timingStartEndRangeList = GetAeneasTimings(requestParameters);
if (timingStartEndRangeList == null)
{
request.ReplyWithText("");
return;
}
string allEndTimesStr = String.Join(" ", timingStartEndRangeList.Select(tuple => tuple.Item2));
Logger.WriteEvent("AudioSegmentationApi.GetForcedAlignmentTimings(): Completed successfully.");
request.ReplyWithText(allEndTimesStr);
}
internal List <Tuple <string, string> > GetAeneasTimings(AutoSegmentRequest requestParameters, out string audioFilenameToSegment, out List <string> fragmentIds)
{
fragmentIds = null;
// The client was supposed to validate this already, but double-check in case something strange happened.
string directoryName = GetAudioDirectory();
audioFilenameToSegment = GetFileNameToSegment(directoryName, requestParameters.audioFilenameBase);
if (string.IsNullOrEmpty(audioFilenameToSegment))
{
Logger.WriteEvent("AudioSegmentationApi.GetAeneasTimings(): No input audio file found.");
ErrorReport.ReportNonFatalMessageWithStackTrace("No audio file found. Please record audio first.");
return(null);
}
IEnumerable <AudioTextFragment> audioTextFragments = requestParameters.audioTextFragments;
string requestedLangCode = requestParameters.lang;
// The client was supposed to validate this already, but double-check in case something strange happened.
// Since this is basically a desperate fallback that shouldn't ever happen we won't try to make the message
// contain a hot link here. That code is in Typescript.
string message;
if (!AreAutoSegmentDependenciesMet(out message))
{
var localizedFormatString = L10NSharp.LocalizationManager.GetString("EditTab.Toolbox.TalkingBook.MissingDependency",
"To split recordings into sentences, first install this {0} system.",
"The placeholder {0} will be replaced with the dependency that needs to be installed.");
ErrorReport.ReportNonFatalMessageWithStackTrace(string.Format(localizedFormatString, message));
return(null);
}
// When using TTS overrides, there's no Aeneas error message that tells us if the language is unsupported.
// Therefore, we explicitly test if the language is supported by the dependency (eSpeak) before getting started.
string stdOut;
string stdErr;
string langCode = GetBestSupportedLanguage(requestedLangCode, out stdOut, out stdErr);
if (string.IsNullOrEmpty(langCode))
{
// FYI: The error message is expected to be in stdError with an empty stdOut, but I included both just in case.
Logger.WriteEvent("AudioSegmentationApi.GetAeneasTimings(): eSpeak error.");
ErrorReport.ReportNonFatalMessageWithStackTrace($"eSpeak error: {stdOut}\n{stdErr}");
return(null);
}
Logger.WriteMinorEvent($"AudioSegmentationApi.GetAeneasTimings(): Attempting to segment with langCode={langCode}");
string textFragmentsFilename = Path.Combine(directoryName, $"{requestParameters.audioFilenameBase}_fragments.txt");
string audioTimingsFilename = Path.Combine(directoryName, $"{requestParameters.audioFilenameBase}_timings.{kTimingsOutputFormat}");
// Clean up the fragments
audioTextFragments = audioTextFragments.Where(obj => !String.IsNullOrWhiteSpace(obj.fragmentText)); // Remove entries containing only whitespace
var fragmentList = audioTextFragments.Select(obj => TextUtils.TrimEndNewlines(obj.fragmentText));
if (langCode != requestedLangCode)
{
string collectionPath = _bookSelection.CurrentSelection.CollectionSettings.FolderPath;
string orthographyConversionMappingPath = Path.Combine(collectionPath, $"convert_{requestedLangCode}_to_{langCode}.txt");
if (File.Exists(orthographyConversionMappingPath))
{
fragmentList = ApplyOrthographyConversion(fragmentList, orthographyConversionMappingPath);
}
}
// Get the GUID filenames (without extension)
fragmentIds = audioTextFragments.Select(obj => obj.id).ToList();
List <Tuple <string, string> > timingStartEndRangeList = null;
try
{
File.WriteAllLines(textFragmentsFilename, fragmentList);
timingStartEndRangeList = GetSplitStartEndTimings(audioFilenameToSegment, textFragmentsFilename, audioTimingsFilename, langCode);
}
catch (Exception e)
{
ProblemReportApi.ShowProblemDialog(null, e,
"AudioSegmentationApi.GetAeneasTimings(): Exception thrown during split stage", "nonfatal");
return(null);
}
try
{
RobustFile.Delete(textFragmentsFilename);
}
catch (Exception e)
{
// These exceptions are unfortunate but not bad enough that we need to inform the user
Debug.Assert(false, $"Attempted to delete {textFragmentsFilename} but it threw an exception. Message={e.Message}, Stack={e.StackTrace}");
}
try
{
RobustFile.Delete(audioTimingsFilename);
}
catch (Exception e)
{
// These exceptions are unfortunate but not bad enough that we need to inform the user
Debug.Assert(false, $"Attempted to delete {audioTimingsFilename} but it threw an exception. Message={e.Message}, Stack={e.StackTrace}");
}
return(timingStartEndRangeList);
}
