TLDR: Google’s Lumiere AI model, which uses the new diffusion model STUNet, has brought AI video generation closer to realism. Lumiere creates videos in one process instead of using small frames, and it generates 80 frames compared to the 25 frames generated by Stable Video Diffusion. While Lumiere still has some artificiality in its videos, it demonstrates the progress made in AI video generation tools. Lumiere is comparable to other AI video generators like Runway and Pika and offers additional features such as image-to-video generation, stylized generation, cinemagraphs, and inpainting. However, there is a risk of misuse of the technology, and the authors of Lumiere’s scientific paper did not explain how to address this.
Google’s Lumiere AI model, which uses the new diffusion model STUNet, has brought AI video generation closer to realism. Lumiere creates videos in one process instead of using small frames, and it generates 80 frames compared to the 25 frames generated by Stable Video Diffusion.
While Lumiere still has some artificiality in its videos, it demonstrates the progress made in AI video generation tools. Lumiere is comparable to other AI video generators like Runway and Pika and offers additional features such as image-to-video generation, stylized generation, cinemagraphs, and inpainting.
However, there is a risk of misuse of the technology, and the authors of Lumiere’s scientific paper did not explain how to address this.