Consistent character with Hunyuan and Skyree
In collaboration with @kartel_ai .ai, we developed a pipeline utilizing ComfyUI alongside the Hunyuan and Skyreel models, both specifically designed for video production. Our process began with volumetric capture, followed by training a LoRA model directly on video data. This allowed us to generate video sequences where a person maintains consistency across multiple shots. The objective is to preserve identity and style over time, pushing the boundaries of AI-generated video. Excited to continue exploring this innovative approach! 🔥