https://liu-yufei.github.io/tags/video-generation/Video GenerationHugo Blox Builder (https://hugoblox.com)en-usWed, 01 Jul 2026 00:00:00 +0000https://liu-yufei.github.io/media/icon_hu17033853472409137288.pnghttps://liu-yufei.github.io/tags/video-generation/https://liu-yufei.github.io/publication/omnipersona/Wed, 01 Jul 2026 00:00:00 +0000https://liu-yufei.github.io/publication/omnipersona/<p><strong>Role: Co-first Author (Equal Contribution)</strong></p> <style> .omni-visuals { margin: 2.75rem auto 3rem; max-width: 1180px; } .omni-visuals h2 { margin: 0 0 1.25rem; font-size: 1.75rem; } .omni-overview { display: block; width: 100%; height: auto; } .omni-example-title { margin-top: 3.5rem !important; } .omni-example { margin: 0; } .omni-example video { display: block; width: 100%; height: auto; border-radius: 0.85rem; background: #fff; box-shadow: 0 1rem 2.5rem rgba(15, 23, 42, 0.14); } .omni-example figcaption { max-width: 58rem; margin: 0.85rem auto 0; color: #64748b; text-align: center; font-size: 0.9rem; line-height: 1.55; } @media (max-width: 720px) { .omni-example figcaption { text-align: left; font-size: 0.8rem; } } </style> <section class="omni-visuals" aria-labelledby="omni-overview-title"> <h2 id="omni-overview-title">Overview</h2> <img class="omni-overview" src="https://liu-yufei.github.io/publication/omnipersona/overview.jpg" alt="OmniPersona framework overview"> <h2 class="omni-example-title">Generation Example</h2> <figure class="omni-example"> <video autoplay muted loop playsinline preload="metadata" poster="/publication/omnipersona/generation-example-poster.jpg" aria-label="OmniPersona generation example with static input conditions on the left and the generated video on the right"> <source src="https://liu-yufei.github.io/publication/omnipersona/generation-example.mp4" type="video/mp4"> </video> <figcaption>The static input conditions and the animated output are combined into one synchronized video, following the presentation layout.</figcaption> </figure> </section> <p>OmniPersona introduces Multimodal Holistic Persona Representation and spatiotemporally decoupled persona injection for unified video creation and editing. I co-led the work and contributed to the holistic identity representation module, combining face identity features with DINOv2 and CLIP features for clothing, hairstyle, and body characteristics.</p> <p>On reference-to-video generation, OmniPersona reaches 79.3% face similarity and 68.7% body similarity, improving over the strongest baseline by 5.2 and 34.2 percentage points, respectively.</p>