video
Rameen
Abdal
Research Scientist at Snap Inc. working on world models, video generation and control. Research Lead, SnapVideo Multi-Modal Training.
I am a Research Scientist at Snap Inc. (Creative Vision Team, Palo Alto Office) working on world models and video generation and control. I am also the Research Lead for SnapVideo Multi-Modal Training, where I maintain the model and provide the research expertise behind it. I was a postdoc at Stanford Computational Imaging Lab, Stanford University working with Prof. Gordon Wetzstein. I completed my Ph.D. in Computer Science at VCC, KAUST, supervised by Prof. Peter Wonka. I worked closely with Prof. Niloy Mitra (UCL and Adobe Research). Before that, I obtained the MS Computer Science degree from KAUST, and the B.Tech degree from National Institute of Technology (NIT), Srinagar, India. My work focuses on world models and controllable video generation, building systems that capture and reproduce motion, interaction, and dynamics in immersive and collaborative environments. Feel free to reach out to me directly.
You can find my full CV here.
Experience and Education
7
Snap Inc.
Research Scientist, Palo Alto
2024 - present
Snap Research
Research Intern, Santa Monica
2022
Adobe Research
Collaborator (Remote), London
2020 - 2022
Postdoc
Stanford University
2023 - 2024
Ph.D. in CS
KAUST
2020 - 2023
MS in CS
KAUST
2018 - 2020
B.Tech in ECE
NIT Srinagar
2014 - 2018
Publications
24* denotes equal contribution
2026
video
EgoPlay: Event-Triggered Video Editing for Egocentric Streams
video
MeshLoom: Feed-Forward Non-Rigid Registration of Mesh Sequences
video
Helix4D: Complex 4D Mesh Generation
NearID: Identity Representation Learning via Near-identity Distractors
ArtifactLens: Hundreds of Labels Are Enough for Artifact Detection with VLMs
Visual Personalization Turing Test
video
Tuning-free Visual Effect Transfer across Videos
2023 - 2025
video
Zero-Shot Dynamic Concept Personalization with Grid-Based LoRA
video
Dynamic Concepts Personalization from Single Videos
Improving the Diffusability of Autoencoders
video
video
2022
video
2021
Barbershop: GAN-based Image Compositing using Segmentation Masks
video
StyleFlow: Attribute-conditioned exploration of stylegan-generated images using conditional continuous normalizing flows
2020
Patents
2Committee and Reviewer
SIGGRAPH ASIA 25, 26 (TPC)SIGGRAPH 26 (TPC)EUROGRAPHICS 24 (IPC)ICML 2024TOGTVCGTPAMIAAAI 22-24NEURIPS 23-25ICLR 24-26CVPR 21-25ICCV 21/23/25ECCV 22/24/26SIGGRAPH 21-25SIGGRAPH ASIA 21-24
Talks
3Stanford Computational Imaging Lab, 2022
EXTRACTING SEMANTICS, GEOMETRY, AND APPEARANCE USING GANS
Stanford University, USA
Rising Stars in AI Symposium (organized by Jurgen Schmidhuber), 2022
EXTRACTING SEMANTICS, GEOMETRY, AND APPEARANCE USING GANS
KAUST, KSA
Adobe Research, 2022
EXTRACTING SEMANTICS, GEOMETRY, AND APPEARANCE USING STYLEGAN
San Jose, USA
Ethics and Social Impact
Generative AI, including personalized video generation, opens real opportunities in education, storytelling, virtual production, and accessibility, but it also carries genuine risk of misuse for deepfakes, identity manipulation, and misleading content, as well as the risk of encoding bias into its outputs. In our work we use celebrity images and video footage strictly under the fair use doctrine, for research, commentary, and analysis only. We do not condone malicious use of these methods, and we advocate for responsible development and deployment.
Contact
Address
Palo Alto, California