The inference code

[MontaEllis]

Great job! I'm now able to achieve promising results. However, the code requires approximately 46GB of GPU memory when running inference on 50 frames. Could you share a script for processing long videos, such as using patches or another method?

[subazinga]

Thank you very much for your feature request! We are indeed working on both the performance and efficiency problems. We will release V2 soon! I will keep this issue open until it's done. Please keep paying attention on this project!