Page 1 of 1

Phonemes embeded as audio file markers, or exported as file

PostPosted: Wed Jul 21, 2010 9:37 am
by owntheweb
Hello,
I have several potential website Flash projects involving characters that talk (e.g. a TV style quiz show). In the example of the quiz show, the host will ask all sorts of questions that change all the time to keep things interesting, moving his/her lips to something generated from a service like voiceforge.com.

The problem there though is there's just audio, no means to move the lips in my animations dynamically. I need to either figure out how to analyze raw audio in Flash for phonemes (ouch), use an expensive dedicated Windows hosted service like Character Server (ouch on wallet for non-profit project), skip lip syncing all together (not willing to go there just yet), or... do something else.

One option that sounds cool is to generate two files from the cepstral engine, the audio, and some kind of transcript of phonems with timing of each. Or maybe embed phonem markers in the audio file itself which could possibly be read by Flash Player 9, and 10 (still need to confirm that).

Otherwise, do you have some ideas to analyze audio files before bringing into Flash manually? I know there's some different software out there. One I'm looking into is Flash Lip Sync: http://www.flashpulse.com/ It's less on the fly, but would probably work with some effort.

Any feedback is much appreciated. Thanks in advance! :)

PostPosted: Wed Jul 21, 2010 1:52 pm
by owntheweb
This page is helpful so far: http://www.honeytechblog.com/12-useful- ... for-linux/

And this ('word / phoneme segmentation kit' looks interesting):
http://julius.sourceforge.jp/en_index.php

PostPosted: Mon Aug 09, 2010 10:00 am
by AdamW
The swift command line that is enclosed with all voice versions up to 5.x can output phoneme information (not necessarily the most compact format, but it's all there). For more information, type

swift

at the command line to see all the options. For the user-in-a-hurry, adding --events should do the trick.

PostPosted: Mon Aug 09, 2010 7:26 pm
by owntheweb
That's wonderful! I feel silly for missing that. It's also a fine format to work with.

You're the best!

Thank you,