Does V12 know about bitrates?

Comments

larry-peter wrote on 8/19/2013, 10:09 AM
John, that 8mm transfer correction is totally amazing. Have you posted anywhere the steps you took to achieve that? Many times I've told clients, "There's only so much I can do, this isn't magic." That frame is pretty close to magic.

There must be some sort of temporal analysis going on to gather the resolution needed for the detail in the porch railing and the texture in the snow. I've seen attempts at this in some "uprezzing" software but the end results were totally bogus.

I'm sure tutorials about your process would be prized on this forum.

johnmeyer wrote on 8/19/2013, 10:31 AM
John, that 8mm transfer correction is totally amazing. Have you posted anywhere the steps you took to achieve that? Many times I've told clients, "There's only so much I can do, this isn't magic." That frame is pretty close to magic. Yes I have, and yes it is.

The main place where film restoration techniques are discussed is this very long thread over in the doom9.org forum:

The power of Avisynth: restoring old 8mm films

I was already doing some small amounts of restoration on movie film -- mostly digital dirt removal -- when I read this post (when it was only two pages long instead of thirty-seven). What intrigued me was this short clip posted by the thread progenitor, "VideoFred" (you've got to click on this link and watch this amazing thirty-nine second clip):

Before/After Vimeo Video Clip from VideoFred

I looked at that and marveled at the sharpness and clarity in the restored clip (make sure to watch it in HD, if your computer and connection support HD from Vimeo). Like you, I too thought it was fake until I downloaded his script and applied it to some 1940s 8mm film I was transferring for a client. I used several test clips for my evaluation, and one of them is the film of the porch in Detroit that you see in that still shot. When I saw that, I knew it was real, and got very excited about the possibilities.

However, I thought that VideoFred's script did a poor job of removing dirt, and also thought that he had too many sharpening steps, often making the residual grain look like hailstones. Also, his script operated at about 1-2 frames per second so it took all day to get results. I therefore undertook a complete re-write, using different functions, and applying what I had learned on my own about removing dust and dirt from film. I posted my alternative somewhere in the middle of that long thread, but eventually all the changes took the conversation so far afield that VideoFred started a new thread:

Capturing and restoring old 8mm films

We exchanged emails before he did this, and I agreed to post my suggestions for places everyone should focus in order to further improve the restoration techniques. I then posted my contributions to that effort, which was my second re-write of his script, and the one I now use several times a month, whenever I am transferring film for a client.

While most film I get has been stored in a can, under reasonable conditions, I sometimes get some really awful stuff. This clip came from a reel that was stored in the open (no can) in a basement, for almost sixty years. I could not remove the dirt with film cleaner and a cotton cloth. So, I had to resort almost entirely to digital techniques. The result is not as fantastic as what VideoFred achieved in the clip above, where he started with pristine film, but in its own way my ability to take something truly awful and make it watchable (which is always my goal) I thought was almost as good:



On first viewing you will mostly see the dirt removal, color correction, and motion stabilization. However, replay the clip (in 480p, full screen, if you can) and start looking at the detail on the side of the building, the boats in the harbor in the background, and the details on the slope in front of the building. It's pretty amazing how many things that are almost completely obscured in the original clip now become quite obvious and easy to see:

I hope that helps answer your question.