Open DevinCheung opened 1 week ago
And the pretrained model "SDv1.5_mv_single_ref_nus" is only for multi-view image generation, not multi-view video generation? Thanks!
Thank you for your recognition.
"SDv1.5_mv_single_ref_nus" is a single-frame implementation of the autoregressive generation algorithm, it does has limitations in preserving temporal consistency. We are actively working on a more continuous version and have already achieved some promising results. For more details, you can refer to our report. We may also consider open-sourcing this part of the code in the future.
Thanks for your awesome work! May I ask when the video-version world dreamer would be released? Thanks!