🧠 Browser GAN Lab
A real neural network. No APIs. No shortcuts. Pure math.
Training Data
📁
Drop images here or click to browse
Resized to 16×16 RGB color
Load Sample Data
Architecture
Generator
z[16]
→
Dense 128 + ReLU
→
Dense 128 + ReLU
→
Sigmoid [16×16×3]
Discriminator
img[16×16×3]
→
Dense 128 + LReLU
→
Dense 128 + LReLU
→
Sigmoid [1]
G: 117760 params | D: 115073 params | Total: 232833
Hyperparameters
Learning Rate
0.0020
Batch Size
8
4
8
16
Latent Dimension
16
8
16
32
Image Size
16×16
8×8 (fast, tiny)
16×16 (default)
24×24 (medium)
32×32 (detailed, slower)
48×48 (high-res, slow)
64×64 (max, very slow)
Training Mode
∞ Infinite
Trains until you hit Stop
How GANs Work ▼
▶ Start Training
■ Stop
↺ Reset
Step 0 / ∞
Awaiting data...
Loss Curves
G Loss
D Loss
Generated Output
New Batch
⬇ Export PNG
Generated images will appear here during training
Evolution Timeline
Timeline will appear during training...
Latent Space Explorer
Train a model first, then explore its latent space