Open molo32 opened 3 years ago
Hi, the problem might be caused partially by the alignment of the reference input and partially by the exaggerated initial expression. The model might find it difficult to locate the eyes given its leaning angles. It would be better if the single input image can be aligned by the key points.
The problem becomes severe if we choose some faces with some strange angles
@Kasumigaoka-Utaha why not read our README first and figure out how to align your own image??
Thx, but the results also seems strange, not like what you have shown in the demo
source has 700 frames aligned so:
el csv :
/content/Talking-Face_PC-AVS/misc/Input/faf/ 700 /content/Talking-Face_PC-AVS/misc/Pose_Source/517600078 160 /content/Talking-Face_PC-AVS/misc/Audio_Source/00015.mp3 /content/Talking-Face_PC-AVS/misc/Mouth_Source/681600002 363 dummy"
how to improve the result?
https://user-images.githubusercontent.com/55426197/115946954-f4911600-a49a-11eb-9d71-7a5b66145d90.mp4