S

stable diffusion

SDXLpy is a modular Python framework for building and executing complex Stable Diffusion XL (SDXL) image generation workflows. It enables text-to-image and image-to-image generation, high-quality upscaling, automated region-based detailing via YOLO detection and SAM segmentation, and flexible YAML-driven pipeline orchestration.

Key features include:

Multi-stage SDXL generation with base and refiner models Intelligent upscaling using RealESRGAN, SwinIR, and tile-based processing Automated detailing with bounding box detection, masking, and inpainting Professional CLI built with Typer, supporting progress tracking, debug modes, and configuration overrides Deterministic generation via explicit seed management Structured logging and comprehensive testing (>80% coverage)

Requires Python 3.10+ with GPU recommended for performance. Install via Poetry (poetry install) or pip. Run workflows with sdxl-cli generate --prompt "Your prompt".

Contributions welcome: Follow PEP 8, add tests, and use conventional commits. Licensed under MIT.

GroundBi Master Application is an AI-powered terrain generation application that creates detailed terrain visuals for Original War. The application uses stable diffusion models with LoRA adaptations to generate high-quality terrain images based on reference images.

Accompanying my master thesis going by the same name.