How many passes to encode DVD

[Guiboche]

Hello,

I was just wondering what might be the optimal amount of passes to encode a file to MPEG2 for DVD.

I have first generation native DV footage, and I want to encode it to DVD with optimal results.

As well, does anyone know how many passes they do to encode professional DVDs?

Thanks

Guiboche

[guns1inger]

Most professional DVDs are encoded from very high quality film scans using hardware encoders.

Back in the real world, where you and I operate on our home video, the answer depends on a number of factors. These include

1. How long is the footage ?
2. How long do you want to put on each disk ?
3. What encoders do you have available to you ?
4. What format will you encode the audio to ?
5 Will you have menus, and how much space will they take up ?

Assuming 1 hour or less to a disk, using 2 channel AC3 audio and minimal menus, a constant bitrate encode at 9000 will get you about the best you can hope for. You can probably squeeze a little more out of the encoder (actually, theoretically, you could push it up to around 9600), however some players have a low threshhold for bitrate.

If you want to fit more on the disk then you need to look at multi-pass encodes. Some encoders will allow you to set the number of passes as high as you want (e.g. CCE), others restrict you to a maximum of two passes (e.g. Procoder). Others have built-in optimisations that are proprietry to that encoder (e.g. Procoder's Mastering mode).

Eventually, you hit a point of diminishing returns. Assume it takes 2 hours to encode a single pass, 2 passes takes 4 hours and gives you 98%, 3 passes takes 6 hours and gives you 98.5%, 4 passes takes 8 hours and gives you 98.75%. The improvements become minutely incremental and invisible to the naked eye.

I have seen bold claims of 9 passes in CCE when backing up a disk that was already of high quality. Over 18 hours. There is no way they could honestly state, with a straight face and gun to their temple, that they could see the difference between 9 passes and 3 passes.

The best way to test it to your own satisfaction is to edit together a small sample of your footage. Chose the bits that encoders hate most - whip pans, zooms, fades (cross or to black), fast action in the frame - you only need 5 - 10 minutes. Calculate the correct bitrate based on the amount of total footage (not the sample), and do some test encodes. I think you will find that in most cases 2 passes will be fine, and for more difficult, 3 passes. After that the returns aren't worth the encode time, PC wear and tear, and running costs.

[Guiboche]

Thanks so much for your reply

The file I will be encoding is 20 minutes long, so I will have LOTS of room on the disc.

But it's good to know that even thought I have 99 passes available to me, it's not imperative that I use them

Guiboche

[Marco33]

Even in the CCE manual they recommend 3-4 passes.

FYI

[lenti_75]

2 passes and as much as bitrate you can

[guns1inger]

CCE is a bit of the odd man out. The first pass, which creates the vaf (video analysis file) doesn't count as an encoding pass, so CCE 3 pass is the equivelent, apparently, of everyone elses 2 pass. I don't quite see the logic, and no-one can give me a decent explaination as to why the vaf pass doesn't count. But going by this, the CCE manual's recommendation of 3 - 4 passes is the same as 2-3 passes in another encoder.

[adam]

Its not that the first vaf pass doesn't count, its just that this is how they choose make their gui operate. If you set it to 3 passes this means 3 passes after the creation of the vaf, which in effect is 4 passes. As long as you understand this, its still operating like any other encoder.

The CCE manual does not suggest that one should use 3-4 passes, rather it says that the point of diminishing returns on most sources will be reached at this point. This doesn't mean that the difference between, say a good 1 pass and a 2 pass encode, isn't still minute.

If you compare a 2 pass encode to a 3 pass one, or a 20 pass one for that matter (yes CCE can do this), I believe there is virtually no difference at all. You can look at their bitrate and quantization curves and clearly see differences, but you can also clearly see that those differences are very slight as well. As for a subjective viewing, I'm with guns1inger. With most sources, there will visibly be no difference unless you are using very constrained bitrate settings.

As for commercial DVDs, they are usually encoded scene by scene with the bitrate set manually. If they were to do it all in one go I'm sure they would do at least 2 passes, probably more just for good measure.

My advice has always remained the same. Do as many passes as you can so long as it finishes by the time you wake up in the morning. As long as you get in at least 2, that's plenty.