PosterCraft is a unified framework for generating high-quality aesthetic posters, addressing precise text rendering, abstract art integration, striking layouts, and stylistic harmony. It is designed for researchers and practitioners in generative AI and graphic design seeking advanced poster creation capabilities.

How It Works

PosterCraft employs a four-stage training workflow. It begins with Text Rendering Optimization for accurate text placement on backgrounds, followed by High-quality Poster Fine-tuning using Region-aware Calibration for artistic integrity. Aesthetic-Text RL then optimizes for higher-order aesthetic trade-offs, and finally, Vision-Language Feedback iteratively refines content and harmony through joint conditioning. This multi-stage approach ensures both fidelity and aesthetic appeal.

Quick Start & Requirements

• Installation: Clone the repository, create a conda environment (python=3.11), and install dependencies (pip install -r requirements.txt).
• Prerequisites: Requires a GPU with BF16 precision support.
• Quick Generation: Run inference.py or inference_offload.py with specified prompts and model paths. A Gradio Web UI is also available via demo_gradio.py.
• Resources: Model weights and datasets are available on HuggingFace.
• Links: Website, Paper, HuggingFace Models, HuggingFace Demo.

Highlighted Details

• Achieves state-of-the-art text recall, f-score, and accuracy, outperforming several open and closed-source models.
• Offers specialized datasets for text rendering, poster curation, preference learning, and vision-language feedback.
• Provides pre-trained model weights for different stages of the generation pipeline.
• Demonstrates successful integration into ComfyUI by a community user.

Maintenance & Community

The project is actively maintained with recent updates in June 2025. Community contributions are highlighted, including a ComfyUI integration. Contact information for authors is provided.

Licensing & Compatibility

The repository does not explicitly state a license. Model weights are available on HuggingFace, which typically uses its own terms of service. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The README does not specify any explicit limitations or known bugs. The project appears to be in an active development phase with recent releases.