我一直在尝试使用面部混合形状在 OpenGL 中实现变形目标动画,但遵循本教程。动画的顶点着色器看起来像这样:
#version 400 core
in vec3 vNeutral;
in vec3 vSmile_L;
in vec3 nNeutral;
in vec3 nSmile_L;
in vec3 vSmile_R;
in vec3 nSmile_R;
uniform float left;
uniform float right;
uniform float top;
uniform float bottom;
uniform float near;
uniform float far;
uniform vec3 cameraPosition;
uniform vec3 lookAtPosition;
uniform vec3 upVector;
uniform vec4 lightPosition;
out vec3 lPos;
out vec3 vPos;
out vec3 vNorm;
uniform vec3 pos;
uniform vec3 size;
uniform mat4 quaternion;
uniform float smile_w;
void main(){
//float smile_l_w = 0.9;
float neutral_w = 1 - 2 * smile_w;
clamp(neutral_w, 0.0, 1.0);
vec3 vPosition = neutral_w * vNeutral + smile_w * vSmile_L + smile_w * vSmile_R;
vec3 vNormal = neutral_w * nNeutral + smile_w * nSmile_L + smile_w * nSmile_R;
//vec3 vPosition = vNeutral + (vSmile_L - vNeutral) * smile_w;
//vec3 vNormal = nNeutral + (nSmile_L - nNeutral) * smile_w;
normalize(vPosition);
normalize(vNormal);
mat4 translate = mat4(1.0, 0.0, 0.0, 0.0,
0.0, 1.0, 0.0, 0.0,
0.0, 0.0, 1.0, 0.0,
pos.x, pos.y, pos.z, 1.0);
mat4 scale = mat4(size.x, 0.0, 0.0, 0.0,
0.0, size.y, 0.0, 0.0,
0.0, 0.0, size.z, 0.0,
0.0, 0.0, 0.0, 1.0);
mat4 model = translate * scale * quaternion;
vec3 n = normalize(cameraPosition - lookAtPosition);
vec3 u = normalize(cross(upVector, n));
vec3 v = cross(n, u);
mat4 view=mat4(u.x,v.x,n.x,0,
u.y,v.y,n.y,0,
u.z,v.z,n.z,0,
dot(-u,cameraPosition),dot(-v,cameraPosition),dot(-n,cameraPosition),1);
mat4 modelView = view * model;
float p11=((2.0*near)/(right-left));
float p31=((right+left)/(right-left));
float p22=((2.0*near)/(top-bottom));
float p32=((top+bottom)/(top-bottom));
float p33=-((far+near)/(far-near));
float p43=-((2.0*far*near)/(far-near));
mat4 projection = mat4(p11, 0, 0, 0,
0, p22, 0, 0,
p31, p32, p33, -1,
0, 0, p43, 0);
//lighting calculation
vec4 vertexInEye = modelView * vec4(vPosition, 1.0);
vec4 lightInEye = view * lightPosition;
vec4 normalInEye = normalize(modelView * vec4(vNormal, 0.0));
lPos = lightInEye.xyz;
vPos = vertexInEye.xyz;
vNorm = normalInEye.xyz;
gl_Position = projection * modelView * vec4(vPosition, 1.0);
}
虽然变形目标动画的算法有效,但我在最终计算的混合形状上丢失了面。动画有点像follow gif。
混合形状是从称为 FaceShift 的无标记面部动画软件导出的。
而且,该算法在一个正常的长方体上完美地工作,它在 Blender 中创建了扭曲的混合形状:
我用于面部动画的混合形状会不会有问题?或者我在顶点着色器中做错了什么?
-------------------------------------------------- - - - - - - 更新 - - - - - - - - - - - - - - - - - - - ---------------------
因此,按照建议,我对顶点着色器进行了所需的更改,并制作了一个新动画,但我仍然得到相同的结果。
这是更新的顶点着色器代码:
#version 400 core
in vec3 vNeutral;
in vec3 vSmile_L;
in vec3 nNeutral;
in vec3 nSmile_L;
in vec3 vSmile_R;
in vec3 nSmile_R;
uniform float left;
uniform float right;
uniform float top;
uniform float bottom;
uniform float near;
uniform float far;
uniform vec3 cameraPosition;
uniform vec3 lookAtPosition;
uniform vec3 upVector;
uniform vec4 lightPosition;
out vec3 lPos;
out vec3 vPos;
out vec3 vNorm;
uniform vec3 pos;
uniform vec3 size;
uniform mat4 quaternion;
uniform float smile_w;
void main(){
float neutral_w = 1.0 - smile_w;
float neutral_f = clamp(neutral_w, 0.0, 1.0);
vec3 vPosition = neutral_f * vNeutral + smile_w/2 * vSmile_L + smile_w/2 * vSmile_R;
vec3 vNormal = neutral_f * nNeutral + smile_w/2 * nSmile_L + smile_w/2 * nSmile_R;
mat4 translate = mat4(1.0, 0.0, 0.0, 0.0,
0.0, 1.0, 0.0, 0.0,
0.0, 0.0, 1.0, 0.0,
pos.x, pos.y, pos.z, 1.0);
mat4 scale = mat4(size.x, 0.0, 0.0, 0.0,
0.0, size.y, 0.0, 0.0,
0.0, 0.0, size.z, 0.0,
0.0, 0.0, 0.0, 1.0);
mat4 model = translate * scale * quaternion;
vec3 n = normalize(cameraPosition - lookAtPosition);
vec3 u = normalize(cross(upVector, n));
vec3 v = cross(n, u);
mat4 view=mat4(u.x,v.x,n.x,0,
u.y,v.y,n.y,0,
u.z,v.z,n.z,0,
dot(-u,cameraPosition),dot(-v,cameraPosition),dot(-n,cameraPosition),1);
mat4 modelView = view * model;
float p11=((2.0*near)/(right-left));
float p31=((right+left)/(right-left));
float p22=((2.0*near)/(top-bottom));
float p32=((top+bottom)/(top-bottom));
float p33=-((far+near)/(far-near));
float p43=-((2.0*far*near)/(far-near));
mat4 projection = mat4(p11, 0, 0, 0,
0, p22, 0, 0,
p31, p32, p33, -1,
0, 0, p43, 0);
//lighting calculation
vec4 vertexInEye = modelView * vec4(vPosition, 1.0);
vec4 lightInEye = view * lightPosition;
vec4 normalInEye = normalize(modelView * vec4(vNormal, 0.0));
lPos = lightInEye.xyz;
vPos = vertexInEye.xyz;
vNorm = normalInEye.xyz;
gl_Position = projection * modelView * vec4(vPosition, 1.0);
}
另外,我的片段着色器看起来像这样。(与之前相比,我只是添加了新的材质设置)
#version 400 core
uniform vec4 lightColor;
uniform vec4 diffuseColor;
in vec3 lPos;
in vec3 vPos;
in vec3 vNorm;
void main(){
//copper like material light settings
vec4 ambient = vec4(0.19125, 0.0735, 0.0225, 1.0);
vec4 diff = vec4(0.7038, 0.27048, 0.0828, 1.0);
vec4 spec = vec4(0.256777, 0.137622, 0.086014, 1.0);
vec3 L = normalize (lPos - vPos);
vec3 N = normalize (vNorm);
vec3 Emissive = normalize(-vPos);
vec3 R = reflect(-L, N);
float dotProd = max(dot(R, Emissive), 0.0);
vec4 specColor = lightColor*spec*pow(dotProd,0.1 * 128);
vec4 diffuse = lightColor * diff * (dot(N, L));
gl_FragColor = ambient + diffuse + specColor;
}
最后是我从更新代码中得到的动画:
如您所见,我在变形目标动画中仍然缺少一些三角形/面。有关该问题的任何更多建议/评论将非常有帮助。再次提前感谢。:)
更新:
所以按照建议,如果dot(vSmile_R, nSmile_R) < 0
得到以下图像结果,我翻转了法线。
此外,我没有从 obj 文件中获取法线,而是尝试计算自己的(面和顶点法线),但仍然得到相同的结果。