Peihao Zhu

Peihao Zhu

Senior Applied Scientist · Adobe Firefly

I am a Senior Applied Scientist on the Adobe Firefly team, where I focus on long-video generation — pushing the visual and temporal consistency, controllability, and efficiency of video foundation models beyond the short-clip regime. Previously I was a Senior Research Scientist on the ByteDance Seed team, where I was a core contributor to video generation foundation models including Seedance and Seaweed.

I completed my Ph.D. at VCC, KAUST, advised by Prof. Peter Wonka. My research centers on generative AI — diffusion models for video generation, image synthesis and editing — and computer vision and graphics more broadly.

Email Scholar GitHub CV

Selected Work

Adobe Firefly Video — Gen6 / Gen6-long

Core Contributor · Video Foundation Models · Jan 2026 – Present

A main contributor to Adobe's next-generation video foundation model, leading Gen6-long from research exploration through model delivery — improving long-video visual and temporal consistency while optimizing training and inference efficiency.

website
MMCORE

MMCORE — Multimodal Image Editing Foundation Model

Core Contributor · Multimodal Image Editing · Jul 2025 – Dec 2025

Drove core model development and large-scale experimentation for MMCORE, a unified multimodal foundation model for high-quality, instruction-guided image editing — contributing across training, evaluation, and research iteration.

paper website
Seedance 1.0

Seedance 1.0 — Video Generation Foundation Model

Core Contributor · Video Foundation Models · Dec 2024 – Jun 2025

A main contributor to the development and large-scale training of Seedance 1.0. Designed and implemented the distributed data-loading system and large-scale VAE-latent and text-embedding precomputation pipelines underlying the model.

paper video website
Seaweed-7B

Seaweed-7B — Video Generation Foundation Model

Founding / Core Contributor · Video Foundation Models · Mar 2024 – Nov 2024

Built Seaweed-7B from its early stage as a main contributor. Designed the VAE for efficient visual-information compression and built the distributed data-loading system for reliable, high-throughput video-model training.

paper video website
AI

TikTok AI-moji

Core Contributor · Generative Avatars · Jul 2023 – Dec 2024

Drove generative-avatar model development for TikTok AI-moji, from research prototyping through product deployment — personalization and generation of stylized avatars from user images at product scale.

blog platform

Publications

MMCORE

MMCORE: MultiModal COnnection with Representation Aligned Latent Embeddings

ByteDance Seed

Technical Report, 2026

paper webpage
Seedance 1.0

Seedance 1.0: Exploring the Boundaries of Video Generation Models

ByteDance Seed

Technical Report, 2025

paper webpage
Seaweed-7B

Seaweed-7B: Cost-Effective Training of Video Generation Foundation Model

ByteDance Seed

Technical Report, 2025

paper webpage
Barbershop

Barbershop: GAN-based Image Compositing using Segmentation Masks

Peihao Zhu, Rameen Abdal, John Femiani, Peter Wonka

ACM Transactions on Graphics (Proc. SIGGRAPH Asia), 2021

paper code video
SEAN

SEAN: Image Synthesis with Semantic Region-Adaptive Normalization

Peihao Zhu, Rameen Abdal, Yipeng Qin, Peter Wonka

Proc. IEEE CVPR (Oral), 2020

paper code video
2023
3DAvatarGAN

3DAvatarGAN: Bridging Domains for Personalized Editable Avatars

Rameen Abdal, Hsin-Ying Lee, Peihao Zhu, Menglei Chai, Aliaksandr Siarohin, Peter Wonka, Sergey Tulyakov

Proc. IEEE CVPR, 2023

paper
2022
HairNet

HairNet: Hairstyle Transfer with Pose Changes

Peihao Zhu, Rameen Abdal, John Femiani, Peter Wonka

Proc. European Conference on Computer Vision (ECCV), 2022

video
CLIP2StyleGAN

CLIP2StyleGAN: Unsupervised Extraction of StyleGAN Edit Directions

Rameen Abdal, Peihao Zhu, John Femiani, Niloy J. Mitra, Peter Wonka

SIGGRAPH Conference Proceedings, 2022

paper
Mind the Gap

Mind the Gap: Domain Gap Control for Single Shot Domain Adaptation for GANs

Peihao Zhu, Rameen Abdal, John Femiani, Peter Wonka

International Conference on Learning Representations (ICLR), 2022

paper code video
2021
Improved StyleGAN Embedding

Improved StyleGAN Embedding: Where are the Good Latents?

Peihao Zhu, Rameen Abdal, Yipeng Qin, John Femiani, Peter Wonka

ArXiv pre-print, 2021

paper video
StyleFlow

StyleFlow: Attribute-conditioned Exploration of StyleGAN-Generated Images using Conditional Continuous Normalizing Flows

Rameen Abdal, Peihao Zhu, Niloy J. Mitra, Peter Wonka

ACM Transactions on Graphics (TOG), 2021

paper code video
Labels4Free

Labels4Free: Unsupervised Segmentation using StyleGAN

Rameen Abdal, Peihao Zhu, Niloy J. Mitra, Peter Wonka

Proc. IEEE ICCV, 2021

paper project video
Flow-Guided Video Inpainting

Flow-Guided Video Inpainting with Scene Templates

Dong Lao, Peihao Zhu, Peter Wonka, Ganesh Sundaramoorthi

Proc. IEEE ICCV, 2021

paper
2020
Large Scale Architecture Asset Extraction

Large Scale Architecture Asset Extraction from Panoramic Imagery

Peihao Zhu, Wamiq Reyaz Para, Anna Fruehstueck, John Femiani, Peter Wonka

IEEE Transactions on Visualization and Computer Graphics (TVCG), 2020

paper code video

Experience

Senior Applied Scientist

Firefly, Adobe · San Jose, CA

Jan 2026 – Present

Senior Research Scientist

Seed Team, ByteDance · San Jose, CA

Sep 2024 – Jan 2026

Research Scientist

Seed Team, ByteDance · San Jose, CA

Jul 2023 – Aug 2024

Research Scientist Intern

Reality Lab, Meta · Burlingame, CA

Oct 2022 – Feb 2023

Research Scientist Intern

Creative Vision Lab, Snap · Los Angeles, CA

May 2022 – Sep 2022

Education

Ph.D. in Computer Science

KAUST, Visual Computing Center · Saudi Arabia

2019 – 2023

M.Sc. in Computer Science

KAUST, Visual Computing Center · Saudi Arabia

2017 – 2019

M.Sc. Candidate in Computer Science

Institute of Automation, Chinese Academy of Sciences · China

2016 – 2017

B.Eng. in Automation · GPA 90/100, Top 5%

Northeastern University · China

2012 – 2016

Awards & Honors