Opendream provides a familiar, Photoshop-like web UI for Stable Diffusion, enabling non-destructive, layered image editing and workflow saving. It targets artists and researchers seeking more control and flexibility in their AI image generation pipelines, offering an extensible platform for integrating new diffusion models and features.

How It Works

Opendream employs a layered, non-destructive editing paradigm, similar to Photoshop, allowing users to build complex image manipulations by stacking operations. Each edit is a distinct layer, preserving original data and enabling deterministic workflow replay. Its architecture is designed for extensibility, allowing new diffusion models and features to be integrated as simple Python functions, exposed via a standardized API with type hints for frontend form generation.

Quick Start & Requirements

• Install via sh ./run_opendream.sh.
• Requires Node.js.
• Setup time is approximately 30 seconds.

Highlighted Details

• Supports non-destructive editing with layers and masks.
• Workflows can be saved as portable JSON files for sharing and reproducibility.
• Extensions are designed to be simple Python functions, facilitating easy integration of new models (e.g., DALL-E, ControlNet, SAM).
• Offers a familiar UI for users accustomed to traditional image editing software.

Maintenance & Community

• Developed as part of Stanford's CS 348K.
• Open to contributions.

Licensing & Compatibility

• Licensed under the MIT License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

Extensions may have their own dependencies that need to be managed separately. The project is presented as a university project, and its long-term maintenance and feature roadmap are not explicitly detailed.