About TeleStyle

TeleStyle is an open-source research initiative dedicated to solving one of the most persistent challenges in generative AI: content-preserving style transfer.

Our team consists of researchers and engineers passionate about pushing the boundaries of what Diffusion Transformers can achieve. By building upon the robust Qwen-Image-Edit foundation, we aim to democratize access to high-fidelity artistic tools that respect the structural integrity of the original content.

Our Mission

We believe that style transfer should not be a compromise between aesthetics and recognizability. Our mission is to provide creators, developers, and researchers with lightweight, efficient models that can apply complex artistic styles to both images and videos without "breaking" the underlying reality of the scene.

The Technology

TeleStyle leverages a unique Curriculum Continual Learning framework, allowing it to learn from both highly curated artistic datasets and diverse real-world samples. This hybrid approach ensures robustness and generalization that prior models struggled to achieve.