Loving it so far! (suggestions)

[xMANAx]

First off, I'm loving FramePack so far. I've been using it pretty much every free minute I've had since you released it. I hope that it will continue to grow and add features as time progresses!

With that said, after using it, there are a few ideas I'd like to toss out.

1.) The ability to input a video (as you currently input an image), have the frames parsed (even if it's only the beginning/end 30-60 frames), and then have the AI add to the video, either at the beginning or end.

2.) For the web app itself to track the checkpoints (where the video generates a new video file after each set of steps is complete), then, if you don't like where the video is headed, you can swap back to a previously generated segment, after which you're allowed to change the prompt in an attempt to get the desired result without doing the entire video over. (this goes hand in hand with my first idea)

I have no idea if either of these two things are possible, but if they are I'd love to see them in the future! Thanks for your hard work!

Quick edit: I also wanted to say that I notice FramePack is leaving a lot of my GPU performance unused. 4080 Super (16GB), and it's sitting at 25-30% utilization during generation no matter where I set the GPU Inference Preserved Memory. Example: I'm currently using 8 of 16GB with 29% utilization.