|
Socoder -> On Topic -> 3d lip-synching |
|
| Thu, 13 Mar 2008, 04:46 | |
HoboBen
|
Hi, Just wondering how this is done - is the sound analysed and the mesh/bones updated on-the-fly, or are models pre-animated with the specified sound? Is this something that can be done easily (read: for free)? If a game came without lip-syncing - maybe just a randomly moving mouth or no facial animations at all, would you feel completely ripped off? (when I saw spiderman 2 for the PS2, especially as they zoomed in on the faces and had them wave their fists emphatically, I felt that they were doing it on the cheap!) -=-=- blog | work | code | more code |
| Thu, 13 Mar 2008, 05:19 | |
Afr0
|
I've no idea how, but my first instinct would be to get a copy of HL2 for cheap (if you don't have it already), and take a look at the Source SDK. Especially since you already have some experience with C++ and OpenGL from some previous projects. Edit: And yes, unless you can find some kind of very good KeyGen or something, you *do* infact need a copy of HL2 to get the source, since you need a Steam account. I've been trying to search for downloads on the net (and I'm not ashamed to admit it either, since I wouldn't actually be able to run the compiled binaries without media), but... no luck!! -=-=- Afr0 Games Project Dollhouse on Github - Please fork! 
|
| Thu, 13 Mar 2008, 05:21 | |
|
HoboBen
|
 you already have some experience with C++ and OpenGL Pffft! I just smashed the ugliest code ever together!  But my friend did buy me a copy of HL2 - I'll take a look at Source SDK. Cheers
-=-=- blog | work | code | more code |
| Thu, 13 Mar 2008, 05:27 | |
|
Afr0
|
Also, here is the documentation: Source SDK Wiki -=-=- Afr0 Games Project Dollhouse on Github - Please fork!
|
| Thu, 13 Mar 2008, 05:56 | |
Jayenkai
|
You know... You could do this without having to deal with source! Remember that source uses insanely complex lip synching, so instead go for the simple Puppet-esque style.. Create a short program, load in the audio, and get ready with the mouse. Click, start audio, record start millisecs.. open=0 Repeat if mousedown(1) and open=0 then open=1:record open millisecs - start.. if mousedown(1)=0 and open=1 then open=0:record close millisecs - start. until audio finished. then save that millisecs data, and whenever you playback the audio, use the data to open+close the chr's mouth. -=-=- ''Load, Next List!''
|
| Thu, 13 Mar 2008, 06:26 | |
|
HoboBen
|
Cool! I was thinking of doing something like that, by taking a sketch of the shape of the sound as seen in audacity and somehow converting it into a data script... Get the human/monkey to do the work instead - nice! Thanks Jay! Hey, there's a thought - Cobra even has a command to slow a sound channel down... I could play it at half speed and get it really accurate
-=-=- blog | work | code | more code |
|
|
| |||
|
| ||||