Omer Bar Tal

icon indicating a website link https://omerbt.github.io/ icon indicating an email [email protected]

icon company Runway

Results 16 comments of Omer Bar Tal

Paint-by-words code

Hi! the the code can be found [here](https://huggingface.co/spaces/weizmannscience/multidiffusion-region-based/blob/main/region_control.py)

When will the code be released?

Thanks! the code for panorama generation is currently available in the repo, as well as integrated into diffusers :) The code for region-based generation will become available very soon!

My custom implemetation in Automatic1111's WebUI

Thank you for implementing MultiDiffusion with the WebUI -- looks great! Regarding larger images -- in the simplest setting of having the same prompt for all views, then almost by...

my test output（Reconstructed video lacks consistency） ？

Inaccurate reconstruction is due to: (i) inaccurate DDIM inversion, (ii) imperfect VAE latent space autoencoder. Interestingly, our method may still overcome issues with the DDIM inversion thanks to our TokenFlow...

Code Release

Thanks! There is no limit in video length. We're working on the release, stay tuned :)

Is only the last layer of the edited frame processed?

Thanks! First: all layers are processed, the reason is that there are skip connections in the architecture, which make it different from just operating on the last layer. Regarding the...