AI-Art utilizes VQGAN + CLIP within Docker containers to generate art. It simplifies art creation for users with GPUs through an easy-to-use web UI, expanding upon Kevin Costa’s original work.
Requirements
- Preferred: NVIDIA GPU
- Fallback: CPU mode (slower performance compared to GPU)
Quick Start
Deploy SDL: Start the deployment process using the SDL here.
Monitor Logs: Watch the logs for model loading status and updates.
GPU Resources
-
256x256 images: Requires 6 GB VRAM
-
512x512 images: Requires 12 GB VRAM
-
1024x1024 images: Requires 24 GB VRAM
Note: For non-square images, ensure total pixel count is within GPU VRAM limits. For example:
- 6 GB VRAM: Up to 384x128 or 128x384
- 12 GB VRAM: Up to 512x512
- 24 GB VRAM: Up to 1024x1024
Non-power of 2 dimensions are acceptable, e.g., 300x100.
Checking GPU Resources
To check GPU usage and ensure optimal performance, refer to the logs and monitor GPU resource allocation.
