Encoding strategy

[nicco]

I have a movie ("MOVIE") whch consist in the main video + the end titles.
I have to encode it in divx, the simplest way to do it is:

"MOVIE" --> divx encoding (900Kbps)--> "MOVIE_DVIX"

No problems here.
By the way I thought I don't mind the end titles to have as a good quality as the main video.
If I encode the whole "MOVIE" stream the end titles may subtract bits from the main video in the encoding process. One solution to avoid this would be to manipulate the .log file with EKG and set a higher modulation to the end-titles' frames.
But this procedure is a quite compless and I had an alternative idea:

1)Split the "MOVIE" into two video files: the main video ("MAIN VIDEO") and the end title ("END TITLES")

2) encode the two video stream separately
"MAIN VIDEO" -->divx encoding (900Kbps)--> "MAIN VIDEO_DIVX"
"END TITLES" -->divx encoding (600Kbps)--> "MAIN VIDEO_DIVX"

3)Put the file together wit a simple "direct-steam copy" process:
"MAIN VIDEO_DIVX"+ "MAIN VIDEO_DIVX"--> direct-steam copy --> "MOVIE_DVIX"
So the end titles have a lower bitrate than the main video.

Is this correct?

:confused:
What do you think?

[obieobieobie]

I think you'd be better off with changing the quant. Use something like Q = 20 or higher. Gordian Knot supports this and it works very well.

I did something like this recently. I had a scene which the codec was having a horrible time with. I tried all the settings I could think of, but the only thing I found that worked was Q=1, which made my 64 second scene about 64 MB, and this with a frame size of 512x288. :)

[jggimi]

Gordian Knot includes this feature; it has the ability to encode the end credit section separately at Q20 or your selected Quantizer level, and combine it with the main video and audio for you.

[nicco]

I think you'd be better off with changing the quant. Use something like Q = 20 or higher. Gordian Knot supports this and it works very well.

Why Gordian Knot? I can use the EKG to do this:confused:

Why do you say encoding the whole movie giving the titles a lower quantization is better than the procedure I wrote above?

[manono]

Hi-

It may or may not be better than your method, but doing it through GKnot is a heck of a lot easier.

However, it also looks as if you can give the end credits lower quality by using the EKG on the whole file. Then you won't have to encode the 2 parts separately and join them again afterwards:

http://www.divx.com/support/divx/guide.php#ekg

Thanks to jggimi for the link.

[jggimi]

It actually does your three steps: "breaks" the film into 2 pieces (well, two .avs scripts), encodes them separately, then combines them with the audio. So it's not better, it's exactly the same.

The difference is that it's an automated process. Doom9's step-by-step guide even includes it as an option.

[nicco]

Thank you very much!:p