Ask HN: Does video processing usually work frame by frame?

4 points by cheeri0 5 years ago | 1 comment

I just watched this excellent video about how to use deepfakelab.

https://duckduckgo.com/?t=ffab&q=deepfacelab+tutorial&iax=videos&ia=videos&iai=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DI0p2jLri2h8

I learned it does the deepfake frame by frame as individual images. I thought it was more aware that it was working on a video than that.

Does video processing always work frame by frame like that or does it ever take into consideration that there are prev/next frames.

davismwfl 5 years ago |

Yes, video is processed frame by frame. Video is essentially just a series of images played back at a specific rate. Different formats & protocols work on how to reduce duplicate sections of the frames, but overall processing is always frame by frame. Given that, there are a number of shortcuts you can apply to frames where you don't need to process each frame 100% separately, but that's more an optimization as you are still processing frames.