Screen.Recording.2024-11-11.at.1.49.49.PM.mov
inspired by https://daylightcomputer.com/ and https://www.sunlit.place/
- this is all just divs with background
- the leaves i sourced from adobe stock photos and downscaled to ~300x500 because i knew i would be blurring it anyways
<div id="blinds">
<div class="shutters">
<div class="shutter"></div>
...
<div class="shutter"></div>
</div>
<div class="vertical">
<div class="bar"></div>
<div class="bar"></div>
</div>
</div>
- gotta make the diffusion effect convincing
- things closer to the wall should be less blurry than things away from the wall
- this utilizes multiple blur layers each with a mask image to constraint its effect zone
#progressive-blur {
position: absolute;
height: 100%;
width: 100%;
}
#progressive-blur>div {
position: absolute;
height: 100%;
width: 100%;
inset: 0;
backdrop-filter: blur(var(--blur-amount));
mask-image: linear-gradient(252deg, transparent, transparent var(--stop1), black var(--stop2), black);
}
#progressive-blur>div:nth-child(1) {
--blur-amount: 6px;
--stop1: 0%;
--stop2: 0%;
}
#progressive-blur>div:nth-child(2) {
--blur-amount: 12px;
--stop1: 40%;
--stop2: 80%;
}
#progressive-blur>div:nth-child(3) {
--blur-amount: 48px;
--stop1: 40%;
--stop2: 70%;
}
#progressive-blur>div:nth-child(4) {
--blur-amount: 96px;
--stop1: 70%;
--stop2: 80%;
}
- created a leaf billowing effect using small amounts of rotate and scale
- to add a bit more dynamism, i used svg filters and the lesser known
feTurbulence
andfeDisplacementMap
tags to add a higher octave of noise to billow individual leaves
<div id="leaves">
<svg style="width: 0; height: 0; position: absolute;">
<defs>
<filter id="wind" x="-20%" y="-20%" width="140%" height="140%">
<feTurbulence type="fractalNoise" numOctaves="2" seed="1">
<animate attributeName="baseFrequency" dur="16s" keyTimes="0;0.33;0.66;1"
values="0.005 0.003;0.01 0.009;0.008 0.004;0.005 0.003" repeatCount="indefinite" />
</feTurbulence>
<feDisplacementMap in="SourceGraphic">
<animate attributeName="scale" dur="20s" keyTimes="0;0.25;0.5;0.75;1" values="45;55;75;55;45"
repeatCount="indefinite" />
</feDisplacementMap>
</filter>
</defs>
</svg>
</div>
#leaves {
position: absolute;
background-size: cover;
background-repeat: no-repeat;
bottom: -20px;
right: -700px;
width: 1600px;
height: 1400px;
background-image: url("./leaves.png");
filter: url(#wind);
animation: billow 8s ease-in-out infinite;
}
@keyframes billow {
0% {
transform: perspective(400px) rotateX(0deg) rotateY(0deg) scale(1);
}
25% {
transform: perspective(400px) rotateX(1deg) rotateY(2deg) scale(1.02);
}
50% {
transform: perspective(400px) rotateX(-4deg) rotateY(-2deg) scale(0.97);
}
75% {
transform: perspective(400px) rotateX(1deg) rotateY(-1deg) scale(1.04);
}
100% {
transform: perspective(400px) rotateX(0deg) rotateY(0deg) scale(1);
}
}
Screen.Recording.2024-11-11.at.2.00.38.PM.mov
- the light/dark transition was a bit too abrupt so i decided to add a transition between the two
- a linear interpolation of colors wasn't super exciting so I decided to create a sunrise/sunset color animation
- these colors are just handpicked from many hours of just looking at sunsets and some experimentation, nothing too fancy here
- theres also a subtle 'glow' that is a very crude approximation of light bouncing off the floor
- finally, skew + rotate + transform didnt get what i wanted because it only gives you affine transforms and i want that nice perspective distortion
- decided to derive a transform matrix for memes, this involved pulling desmos to draw some shapes
- then, we can solve for a 4d perspective transform matrix that
matrix3d
can use in css
matrix3d
values are calculated using the following python script:
import numpy as np
def compute_homography(points_src, points_dst):
if points_src.shape != (4, 2) or points_dst.shape != (4, 2):
raise ValueError("Input arrays must be of shape (4,2)")
A = np.zeros((8, 8))
b = np.zeros(8)
for i in range(4):
x, y = points_src[i]
x_prime, y_prime = points_dst[i]
A[i * 2] = [x, y, 1, 0, 0, 0, -x * x_prime, -y * x_prime]
b[i * 2] = x_prime
A[i * 2 + 1] = [0, 0, 0, x, y, 1, -x * y_prime, -y * y_prime]
b[i * 2 + 1] = y_prime
h = np.linalg.solve(A, b)
H = np.array([
[h[0], h[1], 0, h[2]],
[h[3], h[4], 0, h[5]],
[0.0, 0.0, 1.0, 0.0],
[h[6], h[7], 0.0, 1.0]
])
return H
if __name__ == "__main__":
src = np.array([[-1, 0], [0, 0], [0, -1], [-1, -1]]) * 500
# day
dst = np.array([[-1, -0.1], [0, 0], [0, -1], [-1, -1.7]]) * 500
# night
# dst = np.array([[-1, 0.1], [0, 0], [0, -1], [-1, -1.1]]) * 500
# Flip y-axis in src and dst
src[:, 1] = -src[:, 1]
dst[:, 1] = -dst[:, 1]
H = compute_homography(src, dst)
print("Homography matrix:")
print(H)
print("Homography matrix as matrix3d in column-major order:")
print("transform: matrix3d(")
for col in range(4):
values = ", ".join(f"{H[row, col]:.4f}" for row in range(4))
print(f" {values}," if col < 3 else f" {values}")
print(");")