STARCaster is a spatio-temporal autoregressive video diffusion framework for speech-driven portrait animation and continuous view synthesis. Leveraging strong identity guidance, it further supports subject-consistent yet reference-free talking portraits, allowing recontextualization beyond the constraints of the input image.

• [Dec 2025] Paper released on arXiv.
• Code & pretrained models will be released soon. Stay tuned!

If you find STARCaster useful for your research, please consider citing us:

@article{paraperas2025starcaster,
      title={STARCaster: Spatio-Temporal AutoRegressive Video Diffusion for Identity- and View-Aware Talking Portraits}, 
      author={Paraperas Papantoniou, Foivos and Galanakis, Stathis and Potamias, Rolandos Alexandros and Kainz, Bernhard and Zafeiriou, Stefanos},
      journal={arXiv preprint arXiv:2512.13247},
      year={2025}
}

⭐ If you are interested in this project, please consider starring the repository to receive updates!