Published in IEEE International Conference on Automatic Face and Gesture Recognition (FG)(Oral), 2026
We introduce Polyglot, a single unified diffusion-based architecture for personalized multilingual speech-driven facial animation, using transcript embeddings to encode language-aware information and speaker-style embeddings to capture person-specific habits.
Published in PhD Thesis, University of Florence, 2026
PhD thesis on 3D talking heads, covering advances in realism and generalization for speech-driven 3D facial animation.
Published in international Journal of Computer Vision (IJCV), 2025
We explore two training settings: a supervised one, in which training sequences share a fixed topology within a sequence but any mesh can be animated at test time, and an unsupervised one, which allows effective training with varying mesh structures. Additionally, we highlight the limitations of current evaluation metrics and propose new metrics for better lip-syncing evaluation between speech and facial movements.
Published in IEEE International Conference on Automatic Face and Gesture Recognition (FG)(Oral), 2025
We present a toolkit for a Modularized 3D Face reconstruction Benchmark (M3DFB), where the fundamental components of error computation are segregated and interchangeable, allowing one to quantify the effect of each.
Published in IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2025
We present EmoVOCA, a novel dataset that can be used to train a model to generate emotional 3D talking heads. The proposed method is capable of generating emotional 3D talking heads based on a speech track, an intensity label, an emotion label, and an actor to animate
Published in European Conference on Computer Vision (ECCV), 2024
In this work, we present ScanTalk, a novel framework capable of animating 3D faces in arbitrary topologies including scanned data.
Published in International Conference on Image Analisys and Processing (ICIAP)(Oral), 2023
A novel approach for generating 3D talking heads from raw audio inputs. This method grounds on the idea that speech related movements can be comprehensively and efficiently described by the motion of a few control points located on the movable parts of the face, i.e., landmarks.