New diffusion model enables consistent character generation across video frames

Researchers released a paper detailing temporal consistency in AI video generation. Artists can now maintain character identity without manual inpainting between scenes. This shifts the workflow from frame-by-frame correction to prompt engineering and direction.