Load up the script as you normally would, (if you're not
sure about how to load script then see the script docs or
this tutorial)
With the audio source selected in the mixer (or an explorer
if you feel like looking for it :) run the script. It will
send xsi into a picking session where you are prompted to
select the ppg that holds the sliders you want to animate.
Select the facial_animation ppg in an explorer.
The main autotalk window will popup. fig.2 >>
Replace the current fcurve? - overwrites
any animation that existed previously
Time Range - sets the start and end of
the animation to capture
Key Values - sets a minimum and maximum
value for key frames to take.
Tangents - Auto Compute - sets the tangents
of the keys to standard bell shaped curves.
Tangents - Keys Tangents .... if auto
comp is unchecked then this value is used for both the right
and left tangents values.
Curve Offset time - Offset all the fcurves
created by this number of frames.
Offset Individual.... - provides options
for offsetting the beginning and end keys from the marker,
by default they are set to 2, meaning that if you had a
marker that started on frame 5 and finished on frame 7 then
the first key would be set at 3 and the last would be set
at frame 9.
Offset Max Key - Offsets the middle key
by this number of frames.
Randomise - Provides options for randomising
the max and min keys by a percentage of their original value.
Setup the values as you like and press ok. XSI will now
start making the fcurves. Once it finishes thinking you
can play the animation. If you're not happy with the result
then run the script again and check the Replace
Fcurve option.
The markers I set up are extremely rough so the results
aren't brilliant but you can see how with a little setup
of the markers you can get some nice fast animation.
It's not meant as a final solution, just as a quick way
to rough out the animation. The result is still going to
be limited by the original positioning of the markers, as
clearly demonstrated here :)