TeleStyle ships open image-video style transfer model

This is a practical open release, not just a demo, and it lowers the barrier for teams building custom visual stylization pipelines.

• –Supports both image and video stylization with an emphasis on preserving source content structure.
• –Ships with checkpoints and inference scripts, so developers can run experiments without rebuilding training stacks.
• –Uses curriculum continual learning plus curated and synthetic style triplets to improve style generalization.
• –Adds a video-to-video module aimed at better temporal consistency, a common weakness in stylized video output.