Computer vision · Generative AI

Aviad Dahan

Aviad
Dahan

Generative AI Researcher, Audio-Video Personalization & Generation

PhD Candidate, School of Electrical Engineering, Tel Aviv University

I am a Generative AI researcher focused on audio-video personalization and generation, and a PhD candidate in the School of Electrical Engineering at Tel Aviv University, advised by Prof. Raja Giryes and Prof. Lior Wolf.

I work on generative AI for multimodal audio-video generation and editing, building models that can preserve identity, steer motion, and reshape visual stories.

My research spans personalized audiovisual generation, temporally coherent video models, and image editing. I am also interested in adapting foundation models to high-stakes visual domains, including medical image segmentation.

Experience & collaborations

Publications

SepGen: generation and separation of individual audio stems alongside video
SepGen: Multi-Stem Audio-Video Separation and Generation in a Single Model Aviad Dahan, Rajaei Khatib, Yonatan Bitton, Idan Szpektor, Lior Wolf, Raja Giryes arXiv preprint, 2026
Project & demos arXiv Code Models Dataset
ID-LoRA
ID-LoRA: Identity-Driven Audio-Video Personalization with In-Context LoRA Aviad Dahan*, Moran Yanuka*, Noa Kraicer, Lior Wolf, Raja Giryes European Conference on Computer Vision (ECCV), 2026
Project arXiv Code
Consistent Zero-Shot Video Editing with Multimodal Feature Flow
Consistent Zero-Shot Video Editing with Multimodal Feature Flow Aviad Dahan, Eyal Gomel, Lior Wolf, Raja Giryes IEEE Open Journal of Signal Processing, 2026 · Oral at ICIP 2026
Paper DOI ResearchGate
LiteAttention
LiteAttention: A Temporal Sparse Attention for Diffusion Transformers Dor Shmilovich, Tony Wu, Aviad Dahan, Yuval Domb arXiv preprint, 2025
Project arXiv Code
Video Polyp Segmentation
Video Polyp Segmentation using Implicit Networks Aviad Dahan, Tal Shaharabany, Raja Giryes, Lior Wolf Medical Imaging with Deep Learning (MIDL), 2024
Paper Code
AutoSAM
AutoSAM: Adapting SAM to Medical Images by Overloading the Prompt Encoder Tal Shaharabany, Aviad Dahan, Raja Giryes, Lior Wolf British Machine Vision Conference (BMVC), 2023
arXiv Code