Synthesized Podcasts for your Zune and iPod using SAPI

Jul 09, 2007 at 8:18PM

by Paul-Valentin Borza
Average of 0 out of 5 stars 0 ratings

Sign in to rate

3 comments

Tweet

Share

Share

Description

The guide takes you on your new speech-enabled adventure, as you'll learn to mix text with speech into a simple program that synthesizes XHTML transitional blog posts into the Wave format - for your iPod - using SAPI 5.3 and encodes them into the Windows Media Audio format - for your Zune - using Windows Media Encoder 9 Series API. In the end, you'll be able to make podcasts having just a standard plain text RSS 2.0 feed; plus, you'll be able to play them on at least two of the most popular media players available on the market.

Paul-Valentin Borza

Difficulty: Intermediate

Time Required: 6-10 hours

Cost: Free

Software: Microsoft .NET Framework 3.0 Redistributable Package Windows Media Encoder 9 Series Visual Studio Express C# or VB

Hardware:

Download: Source Code

Introduction

You probably read technology news (e.g. Coding4Fun) daily... Coding4Fun does not deliver podcasts - yet - and many web sites are in a similar situation. Microsoft Anna, the new text-to-speech (TTS) voice in Windows Vista, sounds more human than previous Microsoft voices like Mary, Mike or Sam; as time goes by, voices will become more and more natural and you won't even be able to tell the difference between a synthesized and a real human voice.
Wouldn't it be great to be able to convert blog posts into podcasts? You’ll even get to sync these synthesized podcasts with your Zune and iPod and listen to them wherever you are and whenever you want.

Prerequisites

Please download the Microsoft .NET Framework 3.0 Redistributable Package and install it on your computer (there is no need to install the package on Windows Vista). .NET Framework 3.0 gives you the opportunity to take advantage of the latest Speech API (SAPI 5.3); however, only Windows Vista comes with Microsoft Anna, as the voice is built right into the OS. In case you're running a previous version of Windows, the program will use a previous version of the TTS engine and won't sound as clear and crisp as you would expect.
You'll also need the Windows Media Encoder 9 Series to encode a Wave to a Windows Media Audio.
I hope you are already using Visual Studio Express C# or VB since you’re on the Coding4Fun web site (you're ok with any choice, as the sample is available is both languages).

Devices and Audio Formats

Synthesized podcasts should work on at least these devices:

The Zune supports the Windows Media Audio Standard (.wma) audio format as described on the Microsoft Zune Technical Specifications web page;
The iPod supports the Wave (.wav) audio format as described on the Apple iPod Technical Specifications web page;
The HP iPaq hw6915 comes with Windows Media Player 10 Mobile that supports both the Wave and the Windows Media Audio format;
In case you don't own a portable media player, you can always listen to these audio files on your computer.

Now that you have everything up and ready, let's get started: it's Coding4Fun time!

Getting Started with the Speech Synthesizer

The Speech Synthesizer is initialized with a -1 Rate (values can range between -10 and +10), a 80 Volume (values can range between 0 and 100) and a Female Adult EN-US Voice (Microsoft Anna); please note that a voice will be selected regardless of installed voices. The SpeechSynthesizerRate and SpeechSynthesizerVolume etc. can be easily modified inside "SpeakRssPodcast.exe.config".

Visual C#

41 // Synthesizer

42 this._synthesizer = new SpeechSynthesizer();

43 // Rate

44 this._synthesizer.Rate = this._settings.SpeechSynthesizerRate;

45 // Volume

46 this._synthesizer.Volume = this._settings.SpeechSynthesizerVolume;

47 // Voice

48 this._synthesizer.SelectVoiceByHints(this._settings.SpeechSynthesizerVoiceGender,

49 this._settings.SpeechSynthesizerVoiceAge, this._settings.SpeechSynthesizerVoiceAlternate,

50 this._settings.SpeechSynthesizerVoiceCulture);

51 // Speak Progress

52 this._synthesizer.SpeakProgress += new EventHandler<SpeakProgressEventArgs>(_synthesizer_SpeakProgress);

Visual Basic

38 '' Synthesizer

39 Me._synthesizer = New SpeechSynthesizer()

40 '' Rate

41 Me._synthesizer.Rate = Me._settings.SpeechSynthesizerRate

42 '' Volume

43 Me._synthesizer.Volume = Me._settings.SpeechSynthesizerVolume

44 '' Voice

45 Me._synthesizer.SelectVoiceByHints(Me._settings.SpeechSynthesizerVoiceGender, _

46 Me._settings.SpeechSynthesizerVoiceAge, Me._settings.SpeechSynthesizerVoiceAlternate, _

47 Me._settings.SpeechSynthesizerVoiceCulture)

48 '' Speak Progress

49 AddHandler _synthesizer.SpeakProgress, AddressOf Me._synthesizer_SpeakProgress

Getting Started with the Windows Media Encoder

The Windows Media Encoder is initialized with a default source group that has a single audio source; the source uses the Windows Media Audio 8 for Dial-up Modem (32 Kbps) profile for encoding a voice-only Wave, which is enough to keep the audio quality high and the size small.

Visual C#

54 // Encoder

55 this._encoder = new WMEncoder();

56 // Source Group Collection

57 IWMEncSourceGroupCollection sourceGroupColl = this._encoder.SourceGroupCollection;

58 // Source Group

59 IWMEncSourceGroup sourceGroup = sourceGroupColl.Add(this._settings.WindowsMediaEncoderSourceGroup);

60 // Source

61 this._source = sourceGroup.AddSource(WMENC_SOURCE_TYPE.WMENC_AUDIO);

62 // Profile

63 foreach (IWMEncProfile profile in this._encoder.ProfileCollection)

64 {

65 //Console.WriteLine(profile.Name);

66 if (profile.Name == this._settings.WindowsMediaEncoderProfile)

67 sourceGroup.set_Profile(profile); // Use Profile in VB

68 }

Visual Basic

51 '' Encoder

52 Me._encoder = New WMEncoder()

53 '' Source Group Collection

54 Dim sourceGroupColl As IWMEncSourceGroupCollection = Me._encoder.SourceGroupCollection

55 '' Source Group

56 Dim sourceGroup As IWMEncSourceGroup = sourceGroupColl.Add(Me._settings.WindowsMediaEncoderSourceGroup)

57 '' Source

58 Me._source = sourceGroup.AddSource(WMENC_SOURCE_TYPE.WMENC_AUDIO)

59 '' Profile

60 For Each profile As IWMEncProfile In Me._encoder.ProfileCollection

61 'Console.WriteLine(profile.Name);

62 If (profile.Name = Me._settings.WindowsMediaEncoderProfile) Then

63 sourceGroup.Profile = profile '' Use set_Profile in C#

64 End If

65 Next

Removing XHTML Tags

When you write a post on your blog, you're probably also embedding tags like IMG, DIV and SPAN etc.; such tags should not be spoken as they don't contain relevant audio data. The following piece of code removes these tags from an XHTML document and returns the inner text. You could use regular expressions to achieve the same result as it's just a matter of choice - I like it this way because it's easier to understand.

Visual C#

242 private string buildPlainTextFromXHTML(string xhtmlText)

243 {

244 try

245 {

246 XmlDocument xhtmlDoc = new XmlDocument();

247 xhtmlDoc.LoadXml(@"<!DOCTYPE html PUBLIC '-//W3C//DTD XHTML 1.0 Transitional//EN' 'http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd'>" +

248 "<html xmlns='http://www.w3.org/1999/xhtmlDoc'>" +

249 "<head><title></title></head>" +

250 "<body>" + xhtmlText + "</body>" +

251 "</html>");

252 return xhtmlDoc.InnerText;

253 }

254 catch (Exception ex)

255 {

256 return this._settings.SpeechPromptException;

257 }

258 }

Visual Basic

224 Private Function buildPlainTextFromXHTML(ByVal xhtmlText As String) As String

225 Try

226 Dim xhtmlDoc As XmlDocument = New XmlDocument

227 xhtmlDoc.LoadXml("<!DOCTYPE html PUBLIC '-//W3C//DTD XHTML 1.0 Transitional//EN' 'http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd'>" & _

228 "<html xmlns='http://www.w3.org/1999/xhtmlDoc'>" & _

229 "<head><title></title></head>" & _

230 "<body>" & xhtmlText & "</body>" & _

231 "</html>")

232 Return xhtmlDoc.InnerText

233 Catch ex As Exception

234 Return Me._settings.SpeechPromptException

235 End Try

236 End Function

Building Prompts

What are prompts? A Prompt is an object that describes what the speech synthesizer should say (the text to be spoken), but also how it should say it (emphasis, rate and volume). For more details, please check the PromptBuilder class. The following function gets called every time a post is processed - my sample takes the latest three available posts.

Visual C#

225 private Prompt buildItemPrompt(string itemTitle, DateTime itemPubDate, string itemDescription)

226 {

227 PromptBuilder pb = new PromptBuilder(this._settings.SpeechSynthesizerVoiceCulture);

228 pb.StartParagraph(this._settings.SpeechSynthesizerVoiceCulture);

229 pb.AppendText(String.Format(this._settings.SpeechPromptRssItem,

230 itemTitle, itemPubDate.ToLongDateString(),

231 buildPlainTextFromXHTML(itemDescription)), PromptEmphasis.Moderate);

232 pb.EndParagraph();

233

234 return new Prompt(pb);

235 }

Visual Basic

208 Private Function buildItemPrompt(ByVal itemTitle As String, ByVal itemPubDate As DateTime, ByVal itemDescription As String) As Prompt

209 Dim pb As PromptBuilder = New PromptBuilder(Me._settings.SpeechSynthesizerVoiceCulture)

210 pb.StartParagraph(Me._settings.SpeechSynthesizerVoiceCulture)

211 pb.AppendText(String.Format(Me._settings.SpeechPromptRssItem, _

212 itemTitle, itemPubDate.ToLongDateString(), _

213 buildPlainTextFromXHTML(itemDescription)), PromptEmphasis.Moderate)

214 pb.EndParagraph()

215

216 Return New Prompt(pb)

217 End Function

Speaking Wave Podcasts

I have explicitly told the speech synthesizer to change its output to a file; otherwise, the prompts would have been played to the default audio device (the speakers). It then speaks two prompts: the first prompt contains the name of the channel, and the second prompt contains the title, publish date and description of the post.

Visual C#

126 try

127 {

128 // Start Speak

129 this._synthesizer.SetOutputToWaveFile(waveFullPath);

130 this._synthesizer.Speak(buildChannelPrompt(channelTitle));

131 this._synthesizer.Speak(buildItemPrompt(itemTitle, itemPubDate, itemDescription));

132 // Console

133 Console.WriteLine();

134 Console.WriteLine();

135 // Stop Speak

136 this._synthesizer.SetOutputToNull();

137 if (this._settings.EncodeAsWindowsMediaAudio)

138 encodeWave(waveFullPath);

139 }

140 catch (Exception ex)

141 {

142 // Console

143 Console.ForegroundColor = ConsoleColor.Red;

144 Console.WriteLine(String.Format(this._settings.ConsoleExceptionMessage, ex.Message));

145 Console.ResetColor();

146 }

Visual Basic

117 Try

118 '' Start Speak

119 Me._synthesizer.SetOutputToWaveFile(waveFullPath)

120 Me._synthesizer.Speak(buildChannelPrompt(channelTitle))

121 Me._synthesizer.Speak(buildItemPrompt(itemTitle, itemPubDate, itemDescription))

122 '' Console

123 Console.WriteLine()

124 Console.WriteLine()

125 '' Stop Speak

126 Me._synthesizer.SetOutputToNull()

127 If Me._settings.EncodeAsWindowsMediaAudio Then

128 encodeWave(waveFullPath)

129 End If

130 Catch ex As Exception

131 '' Console

132 Console.ForegroundColor = ConsoleColor.Red

133 Console.WriteLine(String.Format(Me._settings.ConsoleExceptionMessage, ex.Message))

134 Console.ResetColor()

135 End Try

Encoding as Windows Media Audio Podcasts

Encoding from one format to the other is quite simple to do, as seen below. Be careful to always call Flush when the encoder has finished (stopped) encoding, so that you don't leave the newly converted file in an inconsistent state.

Visual C#

154 private void encodeWave(string waveFileName)

155 {

156 try

157 {

158 this._source.SetInput(waveFileName, String.Empty, String.Empty);

159 this._encoder.File.LocalFileName = String.Format(this._settings.WindowsMediaAudioFile, waveFileName);

160 // Start Encode

161 this._encoder.PrepareToEncode(true);

162 this._encoder.Start();

163 // Wait for the encoder to catch up

164 while (this._encoder.RunState != WMENC_ENCODER_STATE.WMENC_ENCODER_STOPPED)

165 { Console.WriteLine(this._encoder.Statistics.EncodingTime); }

166 this._encoder.Flush();

167 }

168 catch (Exception ex)

169 {

170 // Console

171 Console.ForegroundColor = ConsoleColor.Red;

172 Console.WriteLine(String.Format(this._settings.ConsoleExceptionMessage, ex.Message));

173 Console.ResetColor();

174 }

175 }

Visual Basic

143 Private Sub encodeWave(ByVal waveFileName As String)

144 Try

145 Me._source.SetInput(waveFileName, String.Empty, String.Empty)

146 Me._encoder.File.LocalFileName = String.Format(Me._settings.WindowsMediaAudioFile, waveFileName)

147 '' Start Encode

148 Me._encoder.PrepareToEncode(True)

149 Me._encoder.Start()

150 '' Wait for the encoder to catch up

151 While (Me._encoder.RunState <> WMENC_ENCODER_STATE.WMENC_ENCODER_STOPPED)

152 Console.WriteLine(Me._encoder.Statistics.EncodingTime) : End While

153 Me._encoder.Flush()

154 Catch ex As Exception

155 '' Console

156 Console.ForegroundColor = ConsoleColor.Red

157 Console.WriteLine(String.Format(Me._settings.ConsoleExceptionMessage, ex.Message))

158 Console.ResetColor()

159 End Try

160 End Sub

Works on iPod

I own a Black 30GB iPod (Video) and can confirm that the synthesized podcasts sound as expected on the device.

However, I do not own a Zune and can not vouch that it will play these podcasts - it should work!

Demo

To illustrate how these synthesized podcasts sound, please listen to this Wave sample (.wav) or this Windows Media Audio (.wma) that were created using the Microsoft Windows Vista RSS, Windows Vista Editions: What's right for you?. You'll agree that it sounds pretty good... Why don't you give it a try? Change the RssFeedUri to target your blog and run the program; there you go, you have created your own podcasts without a microphone or a recording studio.

Conclusion

You have reached the end of this short guide that showed you how to enhance the world of standard plain text blogs; I hope you have enjoyed reading the article as much as I have enjoyed writing and coding the sample. Please use my Windows Live Messenger Id windowslive@borza.ro to talk with me in case you need further assistance. Thanks to the Microsoft Academic Program Team Romania for support.

Improvements

What should you do to improve what I have already done?

Use the PromptBuilder.AppendAudio member to add extra flavor to the podcasts; more, you could alternate sounds (use melodies that were released under a Creative Commons license).
Use the Regex class to convert HTML posts to plain text.
Encode synthesized audio to the MP3 format; in addition, you should also add tags.

Bio

Paul-Valentin Borza is in its second year of study at the Babes-Bolyai University of Cluj-Napoca, Faculty of Mathematics and Computer Science. Since 2005, he is involved in the Microsoft Student Partners - Microsoft Academic Program Romania. He can be reached through his web site at www.borza.ro.

Tags:

Media, utility, Web, audiovideo

The Discussion

the spark and the pioneer

The guide takes you on your new speech-enabled adventure, as you'll learn to mix text with speech

Last modified Jul 09, 2007 at 9:00PM
Rukmal

Great article! The second link however seems to point to a "Page Not Found" too.
For a while I've been trying to convert SAPI 5.1 output directly to MP3 format and managed to get the LAME ACM codec loaded, but it appears to keep writing Wave files and not really encoding. By any chance, would you have any thoughts to offer?
I was hoping to see your source code to see if I could use SAPI 5.3 to do the same thing, so I'd appreciate if you can point us to the source.
Thanks and keep it up!

Last modified Jun 03, 2008 at 9:00PM
Clint

@Rukmal fixed source code link

Last modified Aug 24, 2010 at 9:00PM