home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Light ROM 4
/
Light ROM 4 - Disc 1.iso
/
text
/
maillist
/
1995
/
0295.doc
/
000081_owner-lightwave-l _Mon Feb 6 09:19:50 1995.msg
< prev
next >
Wrap
Internet Message Format
|
1995-03-19
|
4KB
Return-Path: <owner-lightwave-l>
Received: by netcom23.netcom.com (8.6.9/Netcom)
id IAA05922; Mon, 6 Feb 1995 08:40:45 -0800
Received: from msupa.pa.msu.edu by netcom23.netcom.com (8.6.9/Netcom)
id IAA05910; Mon, 6 Feb 1995 08:40:41 -0800
From: LAMBERT_R@msupa.pa.msu.edu
Date: Mon, 6 Feb 1995 11:41:45 -0500 (EST)
To: lightwave-l@netcom.com
Message-Id: <950206114145.21607a6d@msupa.pa.msu.edu>
Subject: character speech
Sender: owner-lightwave-l@netcom.com
Precedence: bulk
Ernie Wright wrote:
>I haven't written any semi-relevant semi-science in this space in a
>while, so here goes.
>A lot of linguistics texts have sagittal section diagrams (side-view
>line drawings of the nose, lips, tongue and throat) that show how most
>phonemes (language sounds) are created. It would be pretty cool to
>make versions of a talking object model for each phoneme, then write
>a program that could read words, translate them into phoneme sequences,
>and write a scene file with the corresponding chain of morph targets.
>Similar things have been done, but as far as I know they're not readily
>available to animators. If there's someone out there with the time and
>the background (I'm a little short on both), she might want to give
>this some thought.
To Ernie and anyone else who is interested in character speech:
I am currently developing an "affordable" 3D paint application but before
I started this project I did some preliminary work on a program that would
provide the basic features that Ernie has mentioned. I intend to return to
this project as soon as I get my 3D paint package working but realize that
someone may beat me to it! So hear's a freebie if you want it. After
having thought about this for a while I decided that the best approach to
converting speech to character animation was by making a digital recording
of your character's lines. Then design your program such that it analyzes
your sound file in time steps that correspond to single frame steps. The
aim is to determine what phoneme is most likely being pronounced during that
frame and with what intensity. Use this data to write a bones envelope for
the entire scene. I realize that using bones might not be the best
approach and one of the reasons that I haven't all ready written this
program is that I haven't had a chance to use bones yet. However it would
not be to difficult to have the program create morph targets from your
original character. This will give you not only correctly timed mouth
movement but also the proper deformation of the mouth.
I don't know if this is simular to the approach that other people have
used, though I suspect it is! It just seems to me that this would be the
way to approach this problem since you could simply record your voice actor
and then run the application on it and within a few minutes have an accurate
talking character.
If this process sounds a lot like voice recognition then I should point
out that it is really not even close. Since you are only trying to best
guess the most likely phoneme at any frame the worst thing that could happen
is that you get the wrong phoneme sometimes but you can use an averaging
process to weed these mistakes out since the human mouth is incapable of
producing phonemes in a 1/30 of a second. You will always get the mouth to
open at the right time and to the proper width just from the time averaged
intensity for each frame. In fact I'm sure there are people on this list
that could write a program that would open and close a character's mouth in a
few days.
I'm hoping that enough people would express an interest to me for either
an affordable 3D paint package or the voice synchronization software such
that I would feel justified in working on them full time. I think they
would be really useful tools but I honestly don't know how many other
people are really interested in these products. Well I hope to hear from
some of you about these topics.
Rod Lambert lambert_r@msupa.pa.msu.edu