High Def to DVD quality losses?

[Perrone Ford]

All this specuation got my mind wandering last night and this morning. And I decided to try something.

Suppose you could generate video that was clean and with known colors, render that video, and pull it BACK into vegas. Then modify it to match colors with what your started with. The corrections that were necessary would be a fixed quantity that you could save as an effects chain, and you could just drop onto any timeline when you were going to render out. This way you could be sure your colors were correct for the medium you were going to.

So, with that in mind, I created a project that had color bars, gradients, pluges, and porches. And I rendered to MOV (mpeg4), mpg (Mainconcept DVD), WMV, and Cineform. I used 32 bit float, and linear gamma, just like I do with all my projects.

What I got back was interesting.

Going to MOV with the mpeg-4 codec, the video shifted dark by .01. That change in brightnes/contrast put the pluges right back on the money. Color levels were unaffected in both the gradients and the bars. (bitrate 4Mb)

Going to WMV the colors were on the MONEY. It was like looking at generated media. But the video shifted bright by .01. So I corrected that and all was right with the world (bitrate 4mb)

Going to Cineform, the brightness was spot on, as were the colors, but the colors bloomed a bit. VERY interesting. (no options on encode)

Going to Mpeg, the colors bloomed a bit. Especially Magenta and Blue. Cyan was affected a bit, but the other colors seemed untouched. Brightness seemed to drop about .005 but I left them alone. There was some waviness on each line of the waveform when looking at the pluges. I don't know what the codec is doing, but it's strange, and none of the other codecs introduced anything that looked like it.


I did see one thing that shocked the HECK out of me. And it will change the way I work in Vegas from now on when doing correcting work. Changing the preview changes the results dramatically on the WFM and the Vectorscope. Using Best Full is of course accurate. Using Best Auto, is tolerable. But anything less and you really can't rely on the scopes. This really surprised me as I had expected the scopes to remain accurate regardless of my "preview", but this was not the case.

I am very pleased I did this little experiement, but it reinforced what I had been seeing in my work. Staying in 32 bit with linear gamma produced accurate, repeatabe results when going out to MOV, WMV, DVD, and Cineform. I can't speak to the 8-bit workflow since I don't use it. Maybe someone else can redo this with those workflows and see what the results are.

-Perrone

[Walt Noon]

Thanks to everyone for your input! SUPER advice!!! I tried it all, and I learned a lot. You guys are awesome!
I just thought I'd leave a comment with regards to my opinion after "trying it all."

Thanks to David Newman for the suggestion of using computer RGB.
He is exactly right on both counts. First, it does improve the contrast, AND it "fixes" the blacks which definitely improved the washed out look I was getting. BUT, he was also right that the "studio RGB" picture is just plain better otherwise. So, going to computer RGB is not an ideal fix if you're finicky. (This stuff MAKES you finicky!)

So, my humble opinion after much play with this is that the best solution is Perrone Ford's suggestion of test rendering a few seconds to your planned codec, and looking at the results so that you can adjust your picture for that codec.

If I really have time on my hands in the future, I thought I might fiddle with trying to make a preset that boosts the right values to give a good picture when rendered to DVD.
I don't know about you, but in general, that's where my work is going to wind up being seen. At least for now....

Thanks again everyone!

Walt