🔍 Frame Resolution
~40 batches/sec
256×256 in per-pixel mode = 65,536 individual pixels to learn. Training samples random pixel subsets each batch for feasibility. Grab a coffee! ☕
🔬 1×1 Per-Pixel Mode: Every pixel is processed individually. No patches, no compromises. Pure pixel-level neural network magic.
64×64 RGB → 4,096 pixels → 12,288 features (1×1 per-pixel)
Patch Size: 1×1 (per-pixel) 🎯
⚙️ Architecture
🏋️ Training
♾️ Infinite frames — new training data cycles in automatically
Loss—
Epoch—
Time—
Frames—
Batches/s—
Pixels—
✨ Generate
Generated Frames