VRD's re-encoding

[aav]

I'm using the frame-exact editing mode, and I find with surprise that the re-encoding you do doesn't affect my material in any way noticeable to me. So naturally I'm wondering, what does VRD actually do? How destructive is the process? It's not a full re-encode is it?

I've noticed that it shortens the GOPs around the cuts, which is expected, but won't some frame types (B?) need to be derived from future, cut off frames? So you'd look towards those frames and gather the necessary interpolation information or something. I'm also guessing that you could reuse the quantization/DCTs of the original frames... Also, you're creating a new I-frame after each cut?

I'm quite curious, just a brief overview would be very much appreciated. :)

[DanR]

I'm using the frame-exact editing mode, and I find with surprise that the re-encoding you do doesn't affect my material in any way noticeable to me. So naturally I'm wondering, what does VRD actually do? How destructive is the process? It's not a full re-encode is it?When you make a frame accurate cut, VideoReDo goes through a fairly complex process deciding what to do. If the cut happens to fall on a GOP boundary then it does not do a recode. If it the cut falls within a GOP, then the GOP is decoded and a shortened GOP is recoded. Original quantization is preserved through the cuts in order to maintain quality. This is an overly simplified version. You are indeed correct that B-Frames complicate the matter greatly because of their out-of-order sequencing. So depending upon the location of the cut, 2 GOPs might actually be involved in the process.

[Anonymous]

If I use the on screen display for frame cuts, I can definitely hit the I frame if it lies in the right place for editing purposes. I have made the following observation (this is based upon "feel" and not a stopwatch looking at specific frames in the GOP structure), cutting on a P frame seems to frame accurate cut faster than on a B frame. Since most of us are using this to cut commercials out of TV recorded content we have a little leeway as to choosing a P or B frame. Obviously I is the best. Could you make a recommendation as to which would be the next best frame to pick? I think that a short blurb on this in the help would be helpfull to other users

Karl

[DanR]

Hello Karl, Start cuts and end cuts are handled differently.

If the start of a cut begins on an I-Frame no recoding is involved, otherwise there will be some encoding. You can see how much by turning on the debug option in Tools>Options>Advanced Stream. There are a couple lines in the log file telling you how many frames it recoded.

At the end of a cut, if you cut on an I- or P-Frame, no recoding is involved. Recoding only takes place on a B-Frame.