MLVC is a new machine-learning-based video codec developed by Microsoft Research. It significantly reduces bandwidth and storage needs compared to traditional codecs like H.264 and H.265. This improved efficiency enhances streaming and video call quality, especially on less reliable networks. MLVC achieves this by replacing conventional video compression techniques with end-to-end learned neural compression. The codec is designed to run efficiently on modern devices' Neural Processing Units (NPUs), leaving headroom for other system tasks. It has demonstrated real-time performance at resolutions up to 540p on various hardware platforms. MLVC is already being integrated into Microsoft Teams to improve video call quality for users. Microsoft has open-sourced MLVC to encourage community collaboration in its development. The release includes model code, trained weights, and training scripts under an MIT License. The long-term goal is to establish an open, learned video codec ecosystem that outperforms legacy solutions.