bitrate choice for latest DivX codec...

[huesage]

Hello to all,

I would like to backup all my DVDs to DivX on my hard drive. Now I am not interested in getting a specific final filesize per movie, I would like to have a resonable happy medium between filesize and quality.

I was wondering if anyone has any advice on the average bitrate to set for such a project.

Any help and advice is apreciated.

[jggimi]

Hmm... I'd recommend using AutoGK as your encoding suite, which will automatically deal with interlacing or telecining for you, supports batching the encodes, and is highly automated.

For encoding, I'd choose a quality based encode, which is single pass with a fixed quantizer. Since you're not trying to reach a specific filesize, this will give you constant quality rather than constant size. Many would recommend using the default 75%. I would probably use at least 80%, if not higher, but I have a very large monitor. Its up to you, of course, and how you intend to display. I recommend making tests with short pieces of your DVD collection -- a chapter or two, for example -- and pick a fixed quality from testing with your eyes on your display.

NOTE: These will play fine on a PC. But, If you put them on data CD/DVD for playback on a standalone DivX player, you may not be able to use this method. Fixed quantizer encodings do not meet the DivX Certified Theater standard, as the peak bitrate may be beyond the capability of the player.

[huesage]

thank you very much for your reply, i will try a quality based encode, now i was wondering if you would recomend using PSYCHOVISUAL ENHANCEMENT and or the SOURCE PREPROCESSING settings for the DivX codec?

with my tests so far, it seems to actually decrease the quality :-S

[nicco]

I don't think quality-base encode would be the right choiche, but a terrible waste of disk-space (even if tou don't matter).
When you perform a q-based encoding the codec give a constant quantizer (that you choose) to each frame, but some scenes would be very good even at a higher quantizer (=a lower dimension). When you perform a multipass encoding the codec make this choiche for you giving quantizer to each frame based upon the target bitrate you input. So if you imput a relatively high bitrate and use a good pre-processed file with avisynth(denoised, right sized, etc..) you can achieve the same result of a very good encoded-movie with the goal of saving space without decreasing your visual perception of the movie.
I think AutoGK could be a very good automated-tool to do this (I don't use it).
Keep psy OFF, and maybe also b-frames off

P.S.200 posts, thanks to everybody who helped me in this time:D
and a new revisited avatar!

[jggimi]

Well, as you can now see, huesage, there are opinions, but there is no such thing as a single best answer. Nicco and I are both right.

I recommended quality based encoding because it's twice as fast as multipass encoding -- important if you are encoding many films -- and because disk space consumption was not a factor. His opinion is that the time consumed by multipass encoding is worth the ability to give very low Quantizers (such as Q=2) to the scenes that need it, and much higher Quantizers to the scenes that don't. But if your selected bitrate ends up effectively meeting or exceeding the fixed Q you would have chosen, it won't make a significant visible or size difference in the end. The choice, of course, is yours to make.

I recommend the DivX Guide (http://www.divx.com/divx/divxpro/guides/), which describes all of the encoding options you have, and that should help you make informed choices. For example, you asked about PsychoVisual Enhancement; that feature (and the differences between the FAST and SLOW variation) is described in detail beginning on page 56.

[Edits for clarity and spelling]

[huesage]

@jggimi:

I have run a couple of tests at 85% - 80% - 75% and 70%. I also like you beleive that 80% is a good choice. Unfortunately disk space is also an issue, so if I can acheive the same visual quality using two passes and an average bitrate, I will do it without any hesitation as the processing time is well worth the saved disk space to me.

@nicco:

What average bitrate do you recomend I use if running a multipass mode and which app/method would you recomend? I am even whilling to run 3 or more passes if visual quality will be improved (some say that more then 3 passes makes no or little difference).

[nicco]

What average bitrate do you recomend I use
it's not a simple question to answer, the bitrate strongly depends of your video source compressibility (difficult to encode movies require a higer bitrate to achieve the same quality). If you use tools like GK or AutoGK I think there is a compressibility test which guide you through this choiche. But for experience I can tell you that 2000-2500 Kbps is a high bitrate.

which app/method would you recomend You can use all-in-one apps like GKnot or AutoGK, but I'm not usual with dvd-backups (I usually encode DV movies) so maybe someone else can give you more useful informations. There are also guides for that.

am even whilling to run 3 or more passes if visual quality will be improved
In my experience 3-passes are enough (but also 2), for further discussion you can look at this thread http://forum.doom9.org/showthread.php?s=&threadid=84671

[huesage]

i understand that there is no PERFECT or BEST birate, but I am trying to figure out what a good average bitrate would be to use on all movies, so I would have to just find a good bitrate for movies that do not compress very well, and that setting should be fine for movies that compress much better also.

so in your opinion what average bitrate can i stick with in order to get predictable good visual quality throughout all backups?

[jggimi]

Nicco mentioned compressibility testing, done automatically by AGK (when not doing quality based encoding) and semiautomatically by GK.

Compressibility testing takes both content and resolution into account, and we have not discussed resolution, though it is another variable in the puzzle of quality encoding.As you may already know, widescreen movie on DVD will have many fewer pixels per frame than a fullscreen movie, as both will have the same 720 pixels in width prior to resizing.Compressibility testing samples the entire content, usually selecting 5% of all frames. It determines what the maximum bitrate would be with Quantizer=1, and calculates a single value that accounts for bitrate, length, and resolution:value = bits / (pixels X frames)Because the content's maximum compressibility and resolution are variable, there is no one single answer to your question.

Nicco mentioned 2000-2500kbps as a relatively high bitrate. This is usually true when working with DVD sources from modern Hollywood productions in widescreen aspect ratios that have been cleanly transferred from good masters without a lot of edge enhancement or other post-processing between film and DVDbut this is not always the case. Saving Private Ryan, for example, has intentional noise and hand camera work, and you may not like SPR at this sort of bitrate. And if your content is older work, or fullscreen format, in general this guideline may fall short in your eyes.

Trust compressibility testing.

[huesage]

thank you very much for your in depth answer, i really apreciate you having put the time and effort into it.

Thank you.

I have alot of testing to do to see what settings work for me and my standards.

[nicco]

@jggimi

Nicco mentioned 2000-2500kbps as a relatively high bitrate. This is usually true:
when working with DVD sources
from modern Hollywood productions
in widescreen aspect ratios
that have been cleanly transferred from good masters
without a lot of edge enhancement or other post-processing between film and DVD

I usually work with DV-source, which are noisier than DVD, not widescreen (720x576 PAL-DV), and (unfortunately :rolleyes: ) I'm not a modern hollywood producer , but a bitrate value of 2000-2500 is high for me, the bitrate-range I usually use is 900-1200 Kbps.
What is the bitrate-range you usually use, and for what?:)

[jggimi]

It depends, I think, on playback environment..

For example, my very last DivX encode was of a modern Hollywood production of just over 2.5 hours. It was 3-CDs (2.1 GB), and, with two 128kbps audio tracks, the video bitrate was just under 1500kbps. I'd like to redo it at higher bitrate, because while it's not bad on a 15" laptop monitor, it really isn't good enough to me when displayed on my widescreen TV. So in this case, perhaps your 2000-2500 guideline would be fine.

But not always. See this thread titled Ever had a movie you just couldn't rip? (http://forum.doom9.org/showthread.php?s=&threadid=43178) for examples of commercial content that causes problems.

And twice your recommendation may not be enough. I used to do a lot of analog transfers from VHS -- and the last one I did, in September, was 640x480, with 108kbps audio, but it ended up needing 4000kbps video in order to look reasonable on playback on a PC. This particular transfer was of talking heads -- cameras were on tripods, with no motion. I've played it back through an LCD projector onto a screen during a presentation, and, while the audience may not have noticed compression artifacts, *I* certainly did.

[nicco]

but it ended up needing 4000kbps video in order to look reasonable on playback on a PC

WOW!

that's interesting,
sincerely I never exceeded 1200Kbps for my dv-movies (640x464, denoised). They usually look good at their size (640x464), at full screen if I whatch them at at least 50 cm far and in my widwscreen tv. Of couse THERE ARE some encoding artifacts (that many people don't notice but I do), but in most cases they are hardly to notice.


This is the reason I get the 2000-2500 guideline for a high bitrate value.
:)

[jggimi]

That's why, when size is not a consideration, and constant quantizer (quality based) is usable for playback, I often recommend it. But its just a recommendation. In this case, we learned from huesage, after I'd made that initial recommendation, that size was indeed a factor.

And, its why I joined you in recommending compressibility testing, since the outcomes we look for have many variables, and size becomes a decision factor.

The objective of a compressibility test is to select an optimum resolution for a specific file size. File size, of course, determines bitrate.

With GK, the user can adjust file size if the resolution is considered unacceptable, or change AviSynth filters and codec options that affect compressibility, and test again.

With AutoGK, you select the final bitrate before testing begins, and you cannot touch filters or codec settings. Its a highly automated tool. In Advanced Options you could fix image width or set a minimum width, but if you do that, you limit the usefulness of the compression test.

[nicco]

The objective of a compressibility test is to select an optimum resolution for a specific file size. File size, of course, determines bitrate.

You are right, but wat about my dv-movies or your VHS?
They are usually not very long (I think for encoding at 4000Kbps yours are not very long too), so there is no problem to make them fit in 1 cd, in other words when I encode my movies I do not have a "target size" (necessary for dvd-backup) so I can not use (I think) the compr.test.
The only thing I can do is to find a balanced bitrate to have a very good quality with a relative low filesize (the more low is the filesize the more space I save in HD or the more movies I can put in 1CD).
Just an example:
I encoded a dv-movie (640x464) at 1000kbps and it's 78MB with an average quantizer of 6(P)-8(B), then I encoded the same movie @2500Kbps and I get a final size of 140MB with an average quantizer of 3(P)-5(B). Of course the last one is better but no so MUCH better than the first one and it's more than double in size, so my final choiche was the frst one.
If you can suggest me a way to use compr.test (with DV sources) to achieve this target (best result with no waste of space) I'll thank you:)

[jggimi]

You're right. My VHS example was relatively short.

But you ask a question for which I have no good answer: bitrate is easy, but resolution selection is much more of an art than a science. In general, the best practice is to avoid enlarging beyond source size, and, when resizing, use Mod16/32 resolutions. Beyond that, it gets subjective, and there is no "best" answer. You can find many members here who use specific widths regularly; but those width choices tend to be quite variable.

Previewing an AviSynth script will quickly show you the impact of filter choices: cropping, noise reduction, as well as resizing. But it wont' show you what the video will look like after encoding. For complex or difficult content, I usually begin with .avs previewing, but sometimes I test several scenes to see what they will look like after encoding.

Of course, all of this is just my opinion, and I'm just one member here.

[nicco]

Thanks

nicco;)

[manono]

So, back to huesage's original question. I think that it's now been shown that one bitrate does not fit all. And I also believe that jiggmi's first suggestion of doing a quality-based encode using AutoGK was the better one. If it takes up too much space at 80%, then try 65-70%. It's fast and gives even quality for the whole movie. As for whether or not it wastes bits on sections that don't need them, I guess that's debatable, but I don't agree with nicco.

[708145]

What about a tool that really encodes to a certain quality, measured in PSNR or SSIM?
I was dreaming about that for a long time and I'm quite sure that it's not too difficult to do.

Anyone interested and wanting to help?

bis besser,
T0B1A5

[AsTimeGoesBy]

I also have begun to backup all my DVDs as video files. For me what was (and still is) a hard way of self-created experimental work.
There are DVDs i have encoded 4 or 5 times permanently changing settings, bitrate, filters and in meantime even the codec.


Originally posted by huesage
... now i was wondering if you would recomend using PSYCHOVISUAL ENHANCEMENT and or the SOURCE PREPROCESSING settings for the DivX codec?

with my tests so far, it seems to actually decrease the quality :-S Belong my impression, when using DivX v5.21 above 1500kbit/s be very careful with source pre-rpcessing. The 'light pre-processing' of v5.11 was quite ok in most cases, but that of v5.21 works very strong killing tiny details regardless.


Originally posted by huesage
...
so in your opinion what average bitrate can i stick with in order to get predictable good visual quality throughout all backups? If you really need one bitrate eg. for 'interactionless' batch encoding i would start trying around 1500-2000mbit/s. But one average unique bitrate is really not very satifsfying.
As mentioned already here, try at least to respect the resolution, 720x312 or 720x408 makes 20-30% difference of pixels to encode resp. to store within the birate you chose.
Of course there are many other points too affecting bitrate making it very difficult (for me) to estimate the bitrate needs - movements, scene-changes, details, contrast, colors...


A little side-note: For my own taste when watching a DivX file at 720x312 on a 16:9 TV a bitrate 1500kbps is a bit low. Most of those encodings (some of them even were 4pass-EKG-tuned) i have had to replace... :(