arXiv:2511.22425v3 Announce Type: replace 
Abstract: We present WUKONG, a novel training-free framework for high-fidelity textured 3D morphing that takes a pair of source and target prompts (image or text) as input. Unlike conventional methods -- which rely on manual correspondence matching and deformation trajectory estimation (limiting generalization and requiring costly preprocessing) -- WUKONG leverages the generative prior of flow-based transformers to produce high-fidelity 3D transitions with rich texture details. To ensure smooth shape transitions, we exploit the inherent continuity of flow-based generative processes and formulate morphing as an optimal transport barycenter problem. We further introduce a sequential initialization strategy to prevent abrupt geometric distortions and preserve identity coherence. For faithful texture preservation, we propose a similarity-guided semantic consistency mechanism that selectively retains high-frequency details and enables precise control over blending dynamics. This avoids common artifacts like oversmoothing while maintaining semantic fidelity. Extensive quantitative and qualitative evaluations demonstrate that WUKONG significantly outperforms state-of-the-art methods, achieving superior results across diverse geometry and texture variations.

تم تقديم WUKONG كإطار جديد بدون تدريب لتحويلات ثلاثية الأبعاد عالية الدقة مع تفاصيل نسيجية، حيث يستخدم نماذج قائمة على التدفق لإنشاء انتقالات سلسة بين الأشكال ثلاثية الأبعاد بناءً على مطالبات المصدر والهدف. تلغي هذه الطريقة الحاجة إلى المطابقة اليدوية والتقدير لمسار التشويه، وهي قيود شائعة في الطرق التقليدية.

WUKONG se ha presentado como un nuevo marco sin entrenamiento para la morfología 3D texturizada de alta fidelidad, utilizando transformadores basados en flujo para crear transiciones suaves entre formas 3D basadas en indicaciones de origen y destino. Este enfoque elimina la necesidad de coincidencia manual y estimación de trayectoria de deformación, que son limitaciones comunes en los métodos tradicionales.

WUKONG a été présenté comme un nouveau cadre sans entraînement pour le morphing 3D texturé de haute fidélité, utilisant des transformateurs basés sur le flux pour créer des transitions fluides entre des formes 3D basées sur des invites source et cible. Cette approche élimine le besoin de correspondance manuelle et d'estimation de trajectoire de déformation, qui sont des limitations courantes dans les méthodes traditionnelles.

WUKONG has been introduced as a novel training-free framework for high-fidelity textured 3D morphing, utilizing flow-based transformers to create smooth transitions between 3D shapes based on source and target prompts. This approach eliminates the need for manual correspondence matching and deformation trajectory estimation, which are common limitations in traditional methods.

Wukong's 72 Transformations: High-fidelity Textured 3D Morphing via Flow Models

Was this article worth reading? Share it

Humanize AI

SwapAnything.io

Metaflow AI

AI Humanizer

VibeFrame

FLUX AI ART

Ready to build your own newsroom?