GeoV2V: Geometry-Grounded Video Diffusion Model for Driving Scene Generation

A night drive through a film festival lobby: ink-black walls, tight grotesk headlines that fade to gray halfway through, and one violet glow on whatever is ours.

xiyuche.github.io/GeoV2VGeoV2V: Geometry-Grounded Video Diffusion Model for Driving Scene Generation
Screenshot of https://xiyuche.github.io/GeoV2V/, © the authors. Shown for reference. Request removal
Venue
ECCV 2026
Year
2026
First author
Yuchen Xi
Field
video generation
Template
custom (hand-built styles.css + app.js)
Theme
Dark

Checked against the rendered page. The large two-tone 'Change the view. Keep the dynamics.' headline in the first viewport is part of the film's poster frame, not HTML. CSS has late overrides (h1 declared clamp(60px,7vw,100px) renders 44px), so all sizes were probed.

Why it is here

Cinematic dark editorial page: big two-tone headlines, synced side-by-side video comparison with viewpoint toggles, and an illustrated formation diagram. It has a strong identity and confident rhythm.

Overview

Color Palette

Typography

Segoe UI / system UI
GeoV2V: Geometry-Grounded Video Diffusion Model for Driving Scene Gene Abstract · Method · Results · BibTeX — 0123456789
ui-monospace
GeoV2V: Geometry-Grounded Video Diffusion Model for Driving Scene Gene Abstract · Method · Results · BibTeX — 0123456789

Spacing & Shape

Page Modules

Guidelines

More like this