Performance Optimization

Optimize ALICE for speed and efficiency

GPU Optimization

Memory Management

  • Enable auto_unload to free GPU memory when idle
  • Adjust cache_size based on VRAM capacity
  • Use FP16 precision for faster generation

Batch Processing

Generate multiple images simultaneously for efficiency. Configure max concurrent generations in settings.

Model Selection

  • Fastest: SD Turbo models (1-4 steps)
  • Balanced: SD 1.5 models (20-30 steps)
  • Quality: SDXL models (30-50 steps)

Resolution Guidelines

  • Match resolution to model native size
  • Higher resolution = more VRAM + slower
  • Use lower resolution for testing

Hardware Recommendations

  • Minimum: 8GB VRAM (SD 1.5)
  • Recommended: 12GB+ VRAM (SDXL)
  • Optimal: 24GB+ VRAM (multiple models cached)

Next Steps