private AVSpeechUtterance GetSpeechUtterance(string text, CrossLocale? crossLocale, float? pitch, float? speakRate, float? volume)
{
AVSpeechUtterance speechUtterance;
var voice = GetVoiceForLocaleLanguage(crossLocale);
if (UIDevice.CurrentDevice.CheckSystemVersion(8, 0))
{
speechUtterance = new AVSpeechUtterance(" ");
speechUtterance.Voice = voice;
}
else
{
speakRate = NormalizeSpeakRate(speakRate);
volume = NormalizeVolume(volume);
pitch = NormalizePitch(pitch);
speechUtterance = new AVSpeechUtterance(text)
{
Rate = speakRate.Value,
Voice = voice,
Volume = volume.Value,
PitchMultiplier = pitch.Value
};
}
return speechUtterance;
}
/// <summary>
/// Speak back text
/// </summary>
/// <param name="text">Text to speak</param>
/// <param name="queue">If you want to chain together speak command or cancel current</param>
/// <param name="crossLocale">Locale of voice</param>
/// <param name="pitch">Pitch of voice</param>
/// <param name="speakRate">Speak Rate of voice (All) (0.0 - 2.0f)</param>
/// <param name="volume">Volume of voice (iOS/WP) (0.0-1.0)</param>
public void Speak(string text, bool queue = false, CrossLocale? crossLocale = null, float? pitch = null, float? speakRate = null, float? volume = null)
{
if (string.IsNullOrWhiteSpace(text))
return;
if (speechSynthesizer == null)
Init();
var speechUtterance = GetSpeechUtterance(text, crossLocale, pitch, speakRate, volume);
SpeakUtterance(queue, speechUtterance);
}
/// <summary>
/// Speak back text
/// </summary>
/// <param name="text">Text to speak</param>
/// <param name="queue">If you want to chain together speak command or cancel current</param>
/// <param name="crossLocale">Locale of voice</param>
/// <param name="pitch">Pitch of voice</param>
/// <param name="speakRate">Speak Rate of voice (All) (0.0 - 2.0f)</param>
/// <param name="volume">Volume of voice (iOS/WP) (0.0-1.0)</param>
public void Speak(string text, bool queue = false, CrossLocale? crossLocale = null, float? pitch = null, float? speakRate = null, float? volume = null)
{
this.text = text;
this.language = crossLocale;
this.pitch = pitch == null ? 1.0f : pitch.Value;
this.speakRate = speakRate == null ? 1.0f : speakRate.Value;
this.queue = queue;
if (textToSpeech == null || !initialized)
{
Init();
}
else
{
Speak();
}
}
/// <summary>
/// Speak back text
/// </summary>
/// <param name="text">Text to speak</param>
/// <param name="queue">If you want to chain together speak command or cancel current</param>
/// <param name="crossLocale">Locale of voice</param>
/// <param name="pitch">Pitch of voice</param>
/// <param name="speakRate">Speak Rate of voice (All) (0.0 - 2.0f)</param>
/// <param name="volume">Volume of voice (iOS/WP) (0.0-1.0)</param>
public async void Speak(string text, bool queue = false, CrossLocale? crossLocale = null, float? pitch = null, float? speakRate = null, float? volume = null)
{
if (string.IsNullOrWhiteSpace(text))
return;
if (speechSynthesizer == null)
Init();
#if !NETFX_CORE
if (!queue)
speechSynthesizer.CancelAll();
#endif
var localCode = string.Empty;
//nothing fancy needed here
if(pitch == null && speakRate == null && volume == null)
{
if(crossLocale.HasValue && !string.IsNullOrWhiteSpace(crossLocale.Value.Language))
{
localCode = crossLocale.Value.Language;
#if NETFX_CORE
var voices = from voice in SpeechSynthesizer.AllVoices
where (voice.Language == localCode
&& voice.Gender.Equals(VoiceGender.Female))
select voice;
speechSynthesizer.Voice =(voices.Any() ? voices.ElementAt(0) : SpeechSynthesizer.DefaultVoice);
#else
var voices = from voice in InstalledVoices.All
where (voice.Language == localCode
&& voice.Gender.Equals(VoiceGender.Female))
select voice;
speechSynthesizer.SetVoice(voices.Any() ? voices.ElementAt(0) : InstalledVoices.Default);
#endif
}
else
{
#if NETFX_CORE
speechSynthesizer.Voice = SpeechSynthesizer.DefaultVoice;
#else
speechSynthesizer.SetVoice(InstalledVoices.Default);
#endif
}
#if NETFX_CORE
try
{
var stream = await speechSynthesizer.SynthesizeTextToStreamAsync(text);
element.SetSource(stream, stream.ContentType);
element.Play();
}
catch(Exception ex)
{
Debug.WriteLine("Unable to playback stream: " + ex);
}
#else
speechSynthesizer.SpeakTextAsync(text);
#endif
return;
}
if(crossLocale.HasValue && !string.IsNullOrWhiteSpace(crossLocale.Value.Language))
{
localCode = crossLocale.Value.Language;
#if NETFX_CORE
var voices = from voice in SpeechSynthesizer.AllVoices
where (voice.Language == localCode
&& voice.Gender.Equals(VoiceGender.Female))
select voice;
#else
var voices = from voice in InstalledVoices.All
where (voice.Language == localCode
&& voice.Gender.Equals(VoiceGender.Female))
select voice;
#endif
if (!voices.Any())
{
#if NETFX_CORE
localCode = SpeechSynthesizer.DefaultVoice.Language;
#else
localCode = InstalledVoices.Default.Language;
#endif
}
}
else
{
#if NETFX_CORE
localCode = SpeechSynthesizer.DefaultVoice.Language;
#else
localCode = InstalledVoices.Default.Language;
#endif
}
if (!volume.HasValue)
volume = 100.0f;
else if (volume.Value > 1.0f)
volume = 100.0f;
else if (volume.Value < 0.0f)
volume = 0.0f;
else
volume = volume.Value * 100.0f;
var pitchProsody = "default";
//var test = "x-low", "low", "medium", "high", "x-high", or "default";
if(!pitch.HasValue)
pitchProsody = "default";
else if(pitch.Value >= 1.6f)
pitchProsody = "x-high";
else if(pitch.Value >= 1.1f)
pitchProsody = "high";
else if(pitch.Value >= .9f)
pitchProsody = "medium";
else if(pitch.Value >= .4f)
pitchProsody = "low";
else
pitchProsody = "x-low";
string ssmlText = "<speak version=\"1.0\" ";
ssmlText += "xmlns=\"http://www.w3.org/2001/10/synthesis\" xml:lang=\"" + localCode + "\">";
ssmlText += "<prosody pitch=\""+pitchProsody+"\" volume=\""+volume.Value +"\" rate=\""+ speakRate.Value+"\" >" + text + "</prosody>";
ssmlText += "</speak>";
#if NETFX_CORE
try{
var stream = await speechSynthesizer.SynthesizeSsmlToStreamAsync(ssmlText);
element.SetSource(stream, stream.ContentType);
element.Play();
}
catch(Exception ex)
{
Debug.WriteLine("Unable to playback stream: " + ex);
}
#else
speechSynthesizer.SpeakSsmlAsync(ssmlText);
#endif
}
/// <summary>
/// Speak back text
/// </summary>
/// <param name="text">Text to speak</param>
/// <param name="queue">If you want to chain together speak command or cancel current</param>
/// <param name="crossLocale">Locale of voice</param>
/// <param name="pitch">Pitch of voice</param>
/// <param name="speakRate">Speak Rate of voice (All) (0.0 - 2.0f)</param>
/// <param name="volume">Volume of voice (iOS/WP) (0.0-1.0)</param>
public void Speak(string text, bool queue = false, CrossLocale? crossLocale = null, float? pitch = null, float? speakRate = null, float? volume = null)
{
if (string.IsNullOrWhiteSpace(text))
return;
if (speechSynthesizer == null)
Init();
var localCode = crossLocale.HasValue &&
!string.IsNullOrWhiteSpace(crossLocale.Value.Language) ?
crossLocale.Value.Language :
AVSpeechSynthesisVoice.CurrentLanguageCode;
pitch = !pitch.HasValue ? 1.0f : pitch;
if (!volume.HasValue)
volume = 1.0f;
else if (volume > 1.0f)
volume = 1.0f;
else if (volume < 0.0f)
volume = 0.0f;
var divid = 4.0f;
if (UIDevice.CurrentDevice.CheckSystemVersion(9, 0))//use default .5f
divid = 2.0f;
else if (UIDevice.CurrentDevice.CheckSystemVersion(8, 0))//use .125f
divid = 8.0f;
else
divid = 4.0f; //use .25f
if (!speakRate.HasValue)
speakRate = AVSpeechUtterance.MaximumSpeechRate / divid; //normal speech, default is fast
else if (speakRate.Value > AVSpeechUtterance.MaximumSpeechRate)
speakRate = AVSpeechUtterance.MaximumSpeechRate;
else if (speakRate.Value < AVSpeechUtterance.MinimumSpeechRate)
speakRate = AVSpeechUtterance.MinimumSpeechRate;
var voice = AVSpeechSynthesisVoice.FromLanguage(localCode);
if(voice == null)
{
Console.WriteLine("Locale not found for voice: " + localCode + " is not valid using default.");
voice = AVSpeechSynthesisVoice.FromLanguage(AVSpeechSynthesisVoice.CurrentLanguageCode);
}
if (UIDevice.CurrentDevice.CheckSystemVersion(8, 0))
{
var dummyUtterance = new AVSpeechUtterance(" ");
dummyUtterance.Voice = voice;
speechSynthesizer.SpeakUtterance(dummyUtterance);
}
var speechUtterance = new AVSpeechUtterance(text)
{
Rate = speakRate.Value,
Voice = voice,
Volume = volume.Value,
PitchMultiplier = pitch.Value
};
if (!queue && speechSynthesizer.Speaking)
speechSynthesizer.StopSpeaking(AVSpeechBoundary.Word);
speechSynthesizer.SpeakUtterance(speechUtterance);
}
private AVSpeechSynthesisVoice GetVoiceForLocaleLanguage(CrossLocale? crossLocale)
{
var localCode = crossLocale.HasValue &&
!string.IsNullOrWhiteSpace(crossLocale.Value.Language) ?
crossLocale.Value.Language :
AVSpeechSynthesisVoice.CurrentLanguageCode;
var voice = AVSpeechSynthesisVoice.FromLanguage(localCode);
if (voice == null)
{
Console.WriteLine("Locale not found for voice: " + localCode + " is not valid. Using default.");
voice = AVSpeechSynthesisVoice.FromLanguage(AVSpeechSynthesisVoice.CurrentLanguageCode);
}
return voice;
}
