public string ProduceSpeechTextItalian(ImageAnalysis imageAnalysis)
{
var resultBuilder = new StringBuilder();
if (imageAnalysis.Faces.Count == 0)
{
resultBuilder.Append("Non vedo persone");
}
else if (imageAnalysis.Faces.Count == 1)
{
var face = imageAnalysis.Faces.First();
resultBuilder.Append(face.Gender == Gender.Female ? "C'è una donna" : "C'è un uomo");
resultBuilder.Append($" che sembra avere { face.Age} anni");
}
else
{
resultBuilder.Append($"Ci sono {imageAnalysis.Faces.Count} persone. Più precisamente: ");
foreach (var face in imageAnalysis.Faces)
{
resultBuilder.Append(face.Gender == Gender.Female ? "una donna" : "un uomo");
resultBuilder.Append($" che sembra avere {face.Age} anni, ");
}
resultBuilder.Append("."); //a little hack
}
var ssml = Modem.BuildSsmlAsync(resultBuilder.ToString(), "it").Result;
return(ssml);
}
public string ProduceSpeechTextEnglish(ImageAnalysis imageAnalysis)
{
var resultBuilder = new StringBuilder();
if (imageAnalysis.Faces.Count == 0)
{
resultBuilder.Append("There are no people around");
}
else if (imageAnalysis.Faces.Count == 1)
{
var face = imageAnalysis.Faces.First();
resultBuilder.Append($"There is one {face.Gender} person of age {face.Age}.");
}
else
{
resultBuilder.Append($"There are {imageAnalysis.Faces.Count} people around. More in detail: ");
foreach (var face in imageAnalysis.Faces)
{
resultBuilder.Append($"a {face.Gender} of age {face.Age}, ");
}
resultBuilder.Append("."); //a little hack
}
var ssml = Modem.BuildSsmlAsync(resultBuilder.ToString(), "en").Result;
return(ssml);
}
private async void btnListen_Click(object sender, RoutedEventArgs e)
{
bool permissionGained = await AudioCapturePermissions.RequestMicrophonePermission();
if (!permissionGained)
{
var ssml = await Modem.BuildSsmlAsync("Could not enable the microphone, please try again.", "en");
var modem = new Modem(AppConfiguration.ModemConfig);
var speech = await modem.ProduceSpeechAsync(ssml);
Play(speech);
}
else
{
txtUtterance.Text = "(listening...)";
var modem = new Modem(AppConfiguration.ModemConfig);
var utterance = await modem.ListenAsync();
switch (utterance.Item1)
{
case ResultReason.RecognizedSpeech:
txtUtterance.Text = utterance.Item2;
break;
case ResultReason.NoMatch:
default:
txtUtterance.Text = "";
var ssml = await Modem.BuildSsmlAsync("Could not understand utterance, please try again.", "en");
var speech = await modem.ProduceSpeechAsync(ssml);
Play(speech);
break;
}
}
}
private async void btnAnalyze_Click(object sender, RoutedEventArgs e)
{
var modem = new Modem(AppConfiguration.ModemConfig);
if (string.IsNullOrWhiteSpace(txtFilePath.Text))
{
var ssml = await Modem.BuildSsmlAsync("Please, either upload or take a new picture.", "en");
var speech = await modem.ProduceSpeechAsync(ssml);
Play(speech);
}
if (string.IsNullOrWhiteSpace(txtUtterance.Text))
{
txtUtterance.Text = "what's up?";
}
string blobUrl = txtFilePath.Text;
var utterance = txtUtterance.Text;
var bitmapImage = new BitmapImage(new Uri(blobUrl, UriKind.Absolute));
imgPhoto.Source = bitmapImage;
var analyzer = new Geordi(AppConfiguration);
Geordi.AnalysisResult response = await analyzer.AnalyzeAsync(blobUrl, utterance);
var result = await modem.ProduceSpeechAsync(response.Result);
Play(result);
txtAnalysisResult.Text = response.Log;
DeletePicture(blobUrl);
txtFilePath.Text = "";
}
public string ProduceSpeechTextItalian(ImageAnalysis imageAnalysis)
{
var ssml = Modem.BuildSsmlAsync(imageAnalysis.Description.Captions.First().Text, "it").Result;
return(ssml);
}
