Koki Nagano

Principal Research Scientist

NVIDIA Research


Hey, welcome to my homepage! My name is Koki Nagano (in Kanji, my first name means “Lighting”, and “Hope”). I work at the intersection of Computer Graphics, Vision and AI. Currently, I am a Principal Research Scientist at NVIDIA. My research focuses on realistic digital humans synthesis, neural media synthesis and trustworthy visual computing including detection and prevention of visual misinformation in collaboration with DARPA. My PhD thesis focused on the topic of high-fidelity 3D human capture (USC, 2017), and I was supervised by Prof. Paul Debevec at USC ICT. Previously I was a Principal Scientist at Pinscreen Inc. During my PhD, I worked for Weta Digital on Hollywood blockbuster movies and for the Meta Reality Labs in Pittsburgh. I got my Bachelor of Engineering from Tokyo Institute of Technology with focus on Environmental Design.

Internship at NVIDIA
We hire research interns. If you are interested in the intersection of Graphics, Vision and GenAI for digital humans, realistic media synthesis (2D/3D), and relevant forgery detection and prevention techniques, please send me an email.


TED Talk

Face-ing Our Realistic Digital Self
TED / TEDxCharlottesville

What You See Is What You GAN

Rendering Every Pixel for High-Fidelity Geometry in 3D GANs

3D Diffusion

GeNVS: Generative Novel View Synthesis with 3D-Aware Diffusion Models
Project, Code

Live 3D Portrait

Real-Time Radiance Fields for Single-Image Portrait View Synthesis

Authenticating Avatars

Avatar Fingerprinting for Authorized Use of Synthetic Talking-Head Videos

Detecting Diffusion Model-Generated Images

On the detection of synthetic images generated by diffusion models
Paper, Code

Neural Portrait Relighting

Lumos: Learning to Relight Portrait Images via a Virtual Light Stage and Synthetic-to-Real Adaptation
Project, Demo

New 3D GAN

Efficient Geometry-aware 3D Generative Adversarial Networks
Project, Code

StyleGAN3 Detectors

Detecting images generated by the latest never-before-seen GAN
Project, Blog

Live Neural Talking Avatar (SIGGRAPH Best-in-show)

I am AI: AI-driven Digital Avatar Made Easy
featured on SIGGRAPH Blog, fxguide

AI Virtual Assistant

AI-Synthesized Avatars: From Real-Time Deepfakes To Photoreal AI Virtual Assistant
featured on Wired, befores and afters

PaGAN Neural Face Renderer

New Real-time Hybrid AI + CGI Face Replacement
featured on fxguide

AI Face Replacement

Advanced Face AI & Neural Rendering
featured on fxguide

World Economic Forum 2020

featured on BBC, CNN, Reuters, Bloomberg


Deep Face Normalization

Live Virtual Try-on

Personalized Avatars for Realtime Virtual Try-on
featured on fxguide.

Detecting Deepfakes

Protecting World Leaders Against Deep Fakes
Paper featured on MIT Tech Review, Al Jazeera, The Verge.

Interactive 3D Hair Modeling

VR Hair Salon for Avatars
featured on befores and afters

World Economic Forum AMNC 2019

The Human Element: Digital Mimicry
watch demo

Netflix Follow This

watch episode on Netflix


paGAN: Real-time Avatars Using Dynamic Textures
featured on LA Times, CBS, CBC, fxguide


Pinscreen Avatars in your Pocket: Mobile paGAN engine and Personalized Gaming
featured on fxguide.

CBC The Fifth State

Deepfake video: The weaponization of fake news - The Fifth Estate
watch episode


Deep Learning-Based Photoreal Avatars for Online Virtual Worlds in iOS
featured on MIT Tech Review, CBS, PBS, fxguide.


Digital Emily 2 Dataset
wikihuman. download dataset


Skin Microstructure Deformation With Displacement Map Convolution
featured on CNET, Fast Company, CG Channel, GIZMODO.

Computer Animation Festival

Skin Stretch: Simulating Dynamic Skin Microgeometry
featured on GIZMODO, The Verge, MOTHERBOARD, fxguide.

Conversations with Life-size Holographic Testimonies

Time-Offset Conversations on a Life-Sized Automultiscopic Projector Array
featured on NBC TODAY, CNN, GIZMODO,

Life-size Holographics 3D Display

An Autostereoscopic Projector Array Optimized for 3D Facial Display
featured on Wall Street Journal, 4gamer.

In the News

Deepfake & Detection
High-Fidelity 3D Capture