🎭 #PoseTalk: Advanced one-shot talking head generation system
• 🗣️ Synthesizes lip-synchronized videos with customizable head poses
• 🔊 Generates poses from both #audio and #text inputs for better motion control
• 🧠 Utilizes Pose Latent Diffusion (#PLD) model to create motion latent from text and audio cues
• 🔬 Addresses loss-imbalance issue with two-stage refinement strategy: #CoarseNet and #RefineNet
• 👁️ Improves lip-synchronization by progressively estimating lip motions
• 🏆 Outperforms state-of-the-art methods in natural head motion synthesis