search

yunik1004 / SAiD

SAiD: Blendshape-based Audio-Driven Speech Animation with Diffusion
https://yunik1004.github.io/SAiD/
Apache License 2.0
70 stars 13 forks source link

generate blendshapes #2

Open duan348733684 opened 7 months ago

duan348733684 commented 7 months ago

Great work, it seems you've only calculated 32 blendshapes, can your method be used to calculate all 52?

yunik1004 commented 7 months ago

Remained blend shapes are related to the eyes (and the tongue - which does not exist in the typical facial mesh). Unfortunately, the deformation transfer algorithm performs bad when transferring the eye blinks (eyeBlinkLeft, eyeBlinkRight). Therefore, you might need to retouch some of the outputs.

SaltedSlark commented 2 weeks ago

@yunik1004 hello,when I try to create my custom blendshape_head by Deformation-Transfer , the position of the eyeballs is so wierd, I have no idea about that. image looking for your reply, thank you so much.

liyuli1997 commented 2 days ago

干得好，您似乎只计算了 32 个混合形状，您的方法是否可以用于计算所有 52 个形状？

您好，你是直接把生成的output_coeffs.csv文件放入unity中吗