FLOAT: Generative Motion Latent Flow Matching for Audio-driven Talking Portrait
This work addresses the problem of generating realistic and expressive talking portraits for applications like virtual avatars or video synthesis, representing an incremental improvement over existing diffusion-based methods.
The paper tackles the challenge of generating temporally consistent and efficient audio-driven talking portrait videos by introducing FLOAT, a flow matching generative model that uses a learned orthogonal motion latent space, resulting in superior visual quality, motion fidelity, and efficiency compared to state-of-the-art methods.
With the rapid advancement of diffusion-based generative models, portrait image animation has achieved remarkable results. However, it still faces challenges in temporally consistent video generation and fast sampling due to its iterative sampling nature. This paper presents FLOAT, an audio-driven talking portrait video generation method based on flow matching generative model. Instead of a pixel-based latent space, we take advantage of a learned orthogonal motion latent space, enabling efficient generation and editing of temporally consistent motion. To achieve this, we introduce a transformer-based vector field predictor with an effective frame-wise conditioning mechanism. Additionally, our method supports speech-driven emotion enhancement, enabling a natural incorporation of expressive motions. Extensive experiments demonstrate that our method outperforms state-of-the-art audio-driven talking portrait methods in terms of visual quality, motion fidelity, and efficiency.