3D facial landmarks from videos jittering

Very excellent work! Thank you so much for your open source.
I tried to extract 3d facial landmarks from a video (one clip of LRS3), but got jittering landmarks.
I use "face-alignment" installed by pip. Should I build from the repo?
Any suggest to stable the landmarks will be appreciated. : )

facial landmarks:
https://youtu.be/QX7Ra3HNs6s
original video:
https://youtu.be/rQlial7BGaI

该 issue 报告了使用 pip 安装的 'face-alignment' 库从视频中提取的 3D 特征点存在抖动问题。首先，检查从仓库构建是否会改变结果。审查视频处理流程，可能是时间平滑或每帧的模型推理问题。查看该库的视频示例脚本，并考虑添加后处理稳定步骤。目标是生成跨帧稳定的特征点。

由索引模型根据 Issue 内容生成。