English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

We introduce TADA, a simple-yet-effective approach that takes textualdescriptions and produces expressive 3D avatars with high-quality geometry andlifelike textures, that can be animated and rendered with traditional graphicspipelines. Existing text-based character generation methods are limited interms of geometry and texture quality, and cannot be realistically animated dueto inconsistent alignment between the geometry and the texture, particularly inthe face region. To overcome these limitations, TADA leverages the synergy of a2D diffusion model and an animatable parametric body model. Specifically, wederive an optimizable high-resolution body model from SMPL-X with 3Ddisplacements and a texture map, and use hierarchical rendering with scoredistillation sampling (SDS) to create high-quality, detailed, holistic 3Davatars from text. To ensure alignment between the geometry and texture, werender normals and RGB images of the generated character and exploit theirlatent embeddings in the SDS training process. We further introduce variousexpression parameters to deform the generated character during training,ensuring that the semantics of our generated character remain consistent withthe original SMPL-X model, resulting in an animatable character. Comprehensiveevaluations demonstrate that TADA significantly surpasses existing approacheson both qualitative and quantitative measures. TADA enables creation oflarge-scale digital character assets that are ready for animation andrendering, while also being easily editable through natural language. The codewill be public for research purposes.

TADA! Text to Animatable Digital Avatars