Publications

All modalities are equal, but video is more equal: Closing the Cross-Attention Gap in Joint Video Generation

Arxiv 2024 Ohad Rahamim, Dvir Samuel, Idan Schwartz, Gal Chechik

[code] [project page] [bibtex]

Video + 3D Motion

Baseline

+ RecCAR

Audio + Video

Baseline

+ RecCAR

Bringing Objects to Life: training-free 4D generation from 3D objects through view consistent noise

Arxiv 2024 Ohad Rahamim, Ori Malca, Dvir Samuel, Gal Chechik

[paper] [code] [bibtex]

Static 3D Mesh

4D Result

Static 3D Mesh

4D Result

Static 3D Mesh

4D Result

Static 3D Mesh

4D Result

Left: Static 3D mesh input | Right: Animated 4D result

Lay-A-Scene: Personalized 3D Object Arrangement Using Text-to-Image Priors

CVPR 2025 Workshop Ohad Rahamim*, Hilit Segev*, Idan Achituve, Yuval Atzmon, Yoni Kasten, Gal Chechik

[paper] [code] [project page] [bibtex]

Pair view: 2D Generated layout image (left) vs. 3D Arranged result (right)

Point-Cloud Completion with Pretrained Text-to-image Diffusion Models

NeurIPS 2023 Yoni Kasten, Ohad Rahamim, Gal Chechik

[project page] [bibtex]

Left: Partial input scan | Middle: Completed surface | Right: Completed surface with input points