Facial customization and lip synching
I understand that lip synching can be done by having various versions of your head model in different speech poses and morphing between them.
I am not sure however, how facial customization (eye distance, nose size, etc.) is done.
Also, how do I combine the two techniques together so I can lip-sync with a facially customized character?
Regards
Rael
I believe you can do the whole thing with morph targets.
Say you''ve got a model of a nose, and two parameters for it, "flared-ness" and "pointyness". Each parameter can range from -1 to 1; by default they''re at 0. That''s your ''control'' position.
For each parameter, you''d have a mesh at each end of the scale (a ''morph target''). The pointyness one for instance - at pointyness = 0, you''d have a very rounded nose mesh stored. Wheras at pointyness = 1, you''d have a very pointy nose mesh stored. Each mesh would have the exact same topology as the control mesh - all you''re doing is moving vertices.
So, you can interpolate (''morph'') between meshes depending on your parameter. The result is, for each parameter, a set of positions a specific group of vertices - the ones that make up the nose, or the eyes, or the legs, or whatever.
Then, I guess you''d do something like average all the positions supplied for a vertex by different parameters. Possibly a weighted average, so that the pointyness of a nose has more influence on the result than the flared-ness.
I''m not sure, though. This is just based on what I''ve played around with in Softimage.
Say you''ve got a model of a nose, and two parameters for it, "flared-ness" and "pointyness". Each parameter can range from -1 to 1; by default they''re at 0. That''s your ''control'' position.
For each parameter, you''d have a mesh at each end of the scale (a ''morph target''). The pointyness one for instance - at pointyness = 0, you''d have a very rounded nose mesh stored. Wheras at pointyness = 1, you''d have a very pointy nose mesh stored. Each mesh would have the exact same topology as the control mesh - all you''re doing is moving vertices.
So, you can interpolate (''morph'') between meshes depending on your parameter. The result is, for each parameter, a set of positions a specific group of vertices - the ones that make up the nose, or the eyes, or the legs, or whatever.
Then, I guess you''d do something like average all the positions supplied for a vertex by different parameters. Possibly a weighted average, so that the pointyness of a nose has more influence on the result than the flared-ness.
I''m not sure, though. This is just based on what I''ve played around with in Softimage.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement