Publication year: 2020
Internet Resource: Please Login to download book
Talking face generation aims to synthesize a sequence of face images and voices of characters corresponding to their identities, it can serve several solutions; such as voice tone, facial expression, with a good lip synchronization. Therefore, we propose virtual news anchor based on real news anchor called Brooke Baldwin, which reads the texts by itself, reports tirelessly 24/7 for anyone in the world. Several methods and algorithms applied in this project using the dataset to simulate the voice, personalize facial expression and natural head poses to present a life-like image instead of a cold robot. Transfer learning and fine-tuning approaches used to serve this project by taking pre-trained models over LRW (Lip reading in the wild) dataset and fine-tune them to achieve our goal.