May 8

Tim O'Reilly

Tim O'Reilly

Listen to O'Reilly Radar

Years ago, I remember a fascinating alpha-geek moment. I had organized a peer-to-peer summit, and one of the people I'd invited was Kevin Lenzo, one of the founders of Festvox, the open source speech synthesis project and Cepstral, a speech synthesis company. One of the other attendees said to him, "Your voice sounds familiar." When Kevin told him who he was, the developer said, "Oh, that's it. I listen to your voice all the time when I pipe IRC channels to festvox so I can listen to them while I code." (Kevin's voice was the basis for one of Cepstral's synthetic voices.)

Well, Allen Noren, who has had an interest in having our books and sites read aloud for years, has just rolled out speech synthesis on our sites and blogs via a partnership with a Swedish company called ReadSpeaker.

Allen talks more about the program on the O'Reilly FYI blog (a good place to keep up on O'Reilly product- and company-related news). But all you really need to know is pretty obvious: click on the "listen" button next to this entry, and a synthetic voice will read it to you. We've just rolled out this feature on Radar. Let us know what you think.

(P.S. A small note on Kevin. He was one of the inspirations for my original idea of the "alpha geek." I kept running into him on so many interesting tracks to the future, that anything Kevin was interested in, I figured was worth paying attention to. In addition to being a pioneer in speech synthesis, he was the creator of a fascinating "infobot" that participated in the Perl IRC channel, one of the first people I heard about climbing on rooftops to set up homebrew WiFi antennas, and founder of the self-organized Yet Another Perl Conference... When I saw that Allen had put up this project with ReadSpeaker, I wondered why we'd never done it with Kevin. A missed opportunity from many years ago.)

tags: backstory  | comments: 12   | Sphere It

Previous  |  Next

0 TrackBacks

TrackBack URL for this entry: http://blogs.oreilly.com/cgi-bin/mt/mt-t.cgi/5470

Comments: 12

  Chris Eich [05.08.07 04:42 PM]

Pretty rough t2s, IMO. (I shudder to think what ReadSpeaker would to with the foregoing abbreviations!)

Specific points:

  • Q&A -> "Q..Annday", alpha-geek -> "Al Fageek", etc. A touch more inter-word delay would help.
  • read -> "reed", even when it should sound like "red". Contextual analysis, anyone?
  • Image ALT text is speechified, unless there isn't any, in which case a gobbledygook file name gets recited.

  Sachin Gaur [05.09.07 01:34 AM]


  Sachin Gaur [05.09.07 01:38 AM]

Can not you guys complete the workflow by doing some thing like a podcast (Mp3 saving of the Speech file).
If i just tick the (RSS )articles and it saves the streams for me as MP3's. May be you give a client side plug-in to dump the stream to mp3.

Anyways just thoughts :-)

  Max Bergstom [05.09.07 03:47 AM]

Really cool!

I found a couple of words that needed some fine tuning so I used the "feedback" link in the player window to submit the word! The more we contribute, the better user experience for all of us!

  Allen Noren [05.09.07 03:54 AM]

Hello Sachin,

One feature of this service is that we can offer podcasts of each blog entry in mp3 format via RSS feeds. We're working on rolling that out now.

  rektide [05.10.07 06:28 PM]

wwwwweiiird, Audrey Tang just mentioned that Larry Wall nearly went to monitor #perl6 with text to speech. mercifully, Tang claims his wife convinced him otherwise.

i stopped following Perl6 probably two years ago. this Tang presentation is fascinating. perl6 in javascript? oh mercy be please!

only 3/4 through, but highly recommended:

  Sachin Gaur [05.16.07 01:47 AM]

Thanks Allen!

  Ben Lobo [05.31.07 02:10 AM]

ReadSpeaker's a great service and it's great to see it working here but it's implementation on most sites, including this one, could be better. When I click the listen link here it resizes my entire browser window down to a tiny box which is a little bit annoying.

Here's an alternative method that's not so disruptive:

  Ben Lobo [06.02.07 02:55 PM]

Thanks for posting my earlier comment about this article, but you didn't include the address of the blog post I was referring to which is: http://benlobo.hobointernet.com/2007/05/readspeaker-without-nasty-page-reloads.html.

  Leo [09.13.07 09:36 PM]

I wish Google would buy this company someday and provide the service for free...:)

  Sachin [12.06.07 11:56 AM]

Hi All,
If this thread is still active, I was thinking of one small idea.
I remember from my usage of festival TTS, that we can smoothen the experience of listening by choosing the appropraite voice(however the typical TTS only comes with male , female voices.)
what if you guys record the different voices(based on geography) and then translating the IP address to the geographical location(can be done using a service like http://www.geobytes.com/IpLocator.htm). based on the origin of request, server picks the appropriate voice(If the IP address is of India then Indian English Accent.)

So people feel a homely Oreilly Radar.


  CrissMiss [04.13.08 11:02 PM]

I really like this feature and noticed that MIT Technology Review is also podcasting their articles using Text to speech. The company they are using, AudioDizer, embeds music and has also the ability to have advertising (msft commercial was playing in one of the files i listened to). Is that a model you are going to use?

Post A Comment:

 (please be patient, comments may take awhile to post)

Type the characters you see in the picture above.