What is the lowest bitrate acceptable to YOU?

[Wensleydale]

How low do you chaps go on a progressive main movie?

Just curious.

[Kedirekin]

Good question. I'm curious what other people are using myself.

In my experience, average bitrate of 3,500 is about the minimum for 24 fps material, down to about 3,000 if there are large black bars (2.35:1 AR).

For full screen 30 fps interlaced extras, 3,800 is rock bottom, and I'm a lot more comfortable above 4,000.

[This might explain why I like transcoders so much. Many DVDs start out with an average bitrate around 4,500; on those DVDs 3,500 corresponds to about 78% in a transcoder].

[influenza]

Well It all depends on the q factor you can get. Do a OPV on it and see for yourself. On a progressive main movie you can go as low as 2200 and still get a q of below 9.

[Wensleydale]

How do I check the q factor?

[influenza]

You let cce create a vaf file (of the opv) and then select multipass and click on the button bit allocation. Fill in min and max and play with the bitrates. Q is showed in lower right of the screen.

[Master Yoda]

Me, i try not to go below about 3800 on on the main movie.But it also depends on the original bitrate.With extras i try not to go below 2000 if the original avg was about 3500 - 4000, i the extras original avg was about 5000 then i try to stay around 3000 - 3500.

Although on animamted stuff you can get away with a lower avg i think.Akira i used a avg of about 3300 - 3400.

I think it really depends person by person.If its a movie with alot of high motion in i may decide i want to try and keep the avg about 4200.If theres hardly any fast motion i may decide to go down to 3500.

[influenza]

well obviously you cannot always choose when doing a full dvd backup (including all extras an menus). 4200 is quite high as most original dvds have an average bitrate of 5000 or so.

most movies will look outstanding at bitrates around 3000 and even below. Progressive stuff that is. Interlaced material requires much more bitrate.

[D3s7]

I personally have done main movies as low as 1900 and w/ 4/5 passes they look good

sometimes you don't have an option if you keep everything

[Trahald]

my *normal* minimum is 2500 for movie (*16:9 or 4:3) and 2200 for (4:3) extras (at 1/2 d1 and interlaced)

my acceptable minimums are 1900 for movie and 1800 for extras.

unfortunately most of my backups need way more than 78% shrinkage

[DDogg]

Originally posted by influenza

You let cce create a vaf file (of the opv) and then select multipass and click on the button bit allocation. Fill in min and max and play with the bitrates. Q is showed in lower right of the screen. @influenza, sometime if you have time, could you do a 1% sample, derive the BR, and see if it agrees with your method above? I believe it should, but it would be good to have an independent verification. To save time there is a derived BR spreadsheet attached in the second post of this (http://forum.doom9.org/showthread.php?s=&threadid=74141) work thread.

/Add: @All - I don't understand when folks mention certain average bitrates in relationship to quality. ABR has nothing to do with quality, which is sorta what influenza said above. The source complexity solely dictates the BR needed for a certain quality, not the other way around and it is uniquely different for each source.

[influenza]

ddogg will do it if I have time this weekend. Ad otherwise you need to be a bit patient as I will be going on vacation for two weeks :D

[DDogg]

Thanks, it should take you less than 5 minutes to do everything :) The vacation sounds fun, anyplace special, or just hanging out?

[Kedirekin]

@DDogg,
I suspect you do understand why people talk about their minimal exceptable ABR. It's human nature to make observations and then generalize based on those observations. Also, I don't agree with your statement that ABR has no relation to quality, but that's an age-old argument that I really don't want to wake.

Personally I'm surprised at the range of minimal acceptable ABRs that people are quoting. Heck, I've had SVCDs that I couldn't stand at an ABR of 2200.

[influenza]

Of course there's a relationship between abr and quality. But that's not what ddogg meant (at least that's what I think ;)). I think what he means is that you cannot say that a certain abr will result in a certain quality.

You cannot say this bitrate will always result in that quality. Just like your statement that some svcd look bad at a BR of 2200. Some others will look excellent at that bitrate.

The question asked here is quite irrelevant though. What is the lowest bitrate acceptable to you is of course very subjective :).

[DDogg]

Originally posted by Kedirekin -
Also, I don't agree with your statement that ABR has no relation to quality, ... A better phasing might have been to the effect that bitrate is only a peripheral byproduct. Bitrate is the container that 'holds' our cargo. Like any container of anything, it is secondary to what is in it. Its shape is dictated by the shape of the contents. Not the other way around. In our case, the "shape" is the compressibility of a given source. Erm, I think my wording was betraying my frustration with the subject, sorry :-oOriginally posted by Kedirekin -
... but that's an age-old argument that I really don't want to wake. Why not? :) Its fun to talk about, especially with folks like you and 'flu'. Anyway, I don't think it is an area of argument that compression characteristics can be quantified in some manner. Compression is the core fundamental of our hobby. Better understanding its base fundamentals are important to enjoying the hobby. At least they are for me. Also, when you can quantify compression characteristics, you can also quantify the effects of various matrices, filters, resolution tweaks, DC precision, and a host of a zillion other things that were members in good standing of the "Subjective Brigade" :-)Originally posted by influenza -
The question asked here is quite irrelevant though. What is the lowest bitrate acceptable to you is of course very subjective". Yep, sure, but how subjective should it be? While I completely agree that what quality somebody is willing to personally accept is personally subjective, it does not have to be a personal mystery. What somebody is willing to accept can be generally quantified in advance. For me, it is a bitrate that can hold a Q of less than 28. That more than fits my display vehicle, which is a 12 year old 32" JVC and helps me to quantify my own subjectivity.

<chuckle> Trying my best to come back on topic, here are two radically different examples I recently ran into which speak to the original question:

1> "Good Fences" - 2 hour, 4:3, true interlaced, plus interlaced interviews and trailers. Spending 3 minutes doing a pre-encode sample showed it required a whopping 4058 Kbps to hold a Q28 on the main movie. The extras were about the same. To get all that on a DVD, and at a personal acceptable quality level, required 1/2D1 resolution for both extras and main. Knowing this up front saved me one heck of a lot aggravation and wasted time. Yeah, kind of an ugly solution, but it looked surprisingly acceptable, although slightly soft. There was no other way to do it, and still be happy with the outcome.

2> "Spirited Away" - 2 hour,16:9, Anim, Progressive. Only requires 1873 Kbps to hold a Q28. Didn't need to spend any more time thinking about it. I knew the space available would allow terrific quality.

So, my reply to the original poster's question is that every source requires a unique bitrate that can provide you with a close approximate to your 'personal level of acceptable quality', or better it. There are several techniques to divine that particular ABR. They are readily available to you, two of which are mentioned in the replies to this post.

[influenza]

Its fun to talk about, especially with folks like you and 'flu'. Anyway

Why me :cool:

[Trahald]

Originally posted by influenza
Why me :cool:

i think he mistakenly thinks you are knowledgeable. (j/k) ;)

anyways.. for me, i wouldnt go 1/2 d1 on my main (extras i have no problem with going 1/2D1). but i do have the option of reencoding the audio(s) smaller.. or even applying some light noise reduction(but not too much because im actually a noise 'fan' ). but for the most part i just live with what bitrate the movie gives me. i was very much into q based decisioning with svcd (tylo's d2sroba) since adding a disk was an easy option. d2sroba being automated didnt hurt either. but ive gotten away with it a bit with dvd and going with bitrate decisioning. for the most part i get away with it (since most movies give you a bitrate that is consistant with an acceptable q)

also, dvds tend to be alot cleaner than they used to be. although some movies are destined to need lots of bitrate to maintain an acceptable Q because there is alot of action and/or purposefully grainy i.e. saving private ryan or band of brothers.

for the most part.. this would benefit my extras since usually i rape them of bitrate for the main movie when maybe the main did not need as much bitrate as i gave it. i'll track my next few projects and see what happens.

[Matthew]

As stated above each source is different, I'm more interested in the q factor values in CCE's advanced settings.

I like these values to be very low, when they get up to around 5 on a consistent basis then I get pretty annoyed.

This is why I never keep extras =)

[DDogg]

@Trahald, you may soon feel right back at home (http://forum.doom9.org/showthread.php?s=&threadid=70717) :)

@Matthew, getting average quantization under 5 is what I shoot for also. Q28 seems to get me there, but I have always been fuzzy on the exact relationship between OPV's unique "Q Value" and the actual average quantization. I guess it would have to be a direct correlation because of the way OPV works, but I never spent the time to directly verify it.

[auenf]

2mbit bitrate can be fine, but it all depends on how much action there is, and how frequent the action is.

use the Q value as a rough indication of how well the video is getting compressed (over 9 means you will most probably have macroblocks).

FWI, i did LOTR1 extended at 2.1mbit ;)

Enf...

[DDogg]

Originally posted by auenf
use the Q value as a rough indication of how well the video is getting compressed (over 9 means you will most probably have macroblocks). auenf, you used the term "Q value". Were you speaking of 'Q-factor' as reported when the VAF is the analyzed, or "Q value" as used in OPV? I think you meant the first? If so, a 'Q-factor' of 9, which I assume is AvgQuant, seems very high to allow a quality encode.

[Trahald]

i believe he means q-factor.. and specifically a 'max' q .. not average.

[Eyes`Only]

For me, it is a bitrate that can hold a Q of less than 28. Wow... I get Q's of around 9 or 10 when I use an ABR of 2500 normally, which btw is my answer to the topic question ie. acceptable bitrate level. Though I've been known to go with 2200-2400 if the DVD has some killer DTS that can't be omitted without a thorough body-washing afterward (I'd feel dirty...)

Using a 36" TV, I have dared many to play the pseudo-Pepsi taste-test with my DVD collection. Without looking at the media, they can't tell the difference. Especially since my extra DVD cases and printed covers are perfect replicas of my originals.

[DDogg]

Originally posted by Eyes`Only
Wow... I get Q's of around 9 or 10 when I use an ABR of 2500 normally, which btw is my answer to the topic question ie. acceptable bitrate level. Though I've been known to go with 2200-2400 if the DVD has some killer DTS that can't be omitted without a thorough body-washing afterward (I'd feel dirty...) Hmm, I wonder if we are speaking of the same thing? Holding an OPV Q-Value of 28 normally gives me an AQuant of 4.50 - 4.80. The ABR to hold that OPV Q value is completely dependent on the needs of the source complexity, and I still remain confused by generalized statements about a certain level of ABR working across any source. Just ain't so. :) It may be 1500 for one source and 3800 for another, although that would be an extreme example for the sake of discussion like Matrix-1 and SPR [to yield the same approximate quality].

[influenza]

Hmm, I wonder if we are speaking of the same thing?

No we're not. If only CCE used the term Q only once :D. Most people here are referencing to the (in)famous Q from the getting best out of CCE robshot guide I guess. The guide states qs below 9 are good doesn't it?

[DDogg]

Most people here are referencing to the (in)famous Q from the getting best out of CCE robshot guide I guess. Oh, I thought that had been obsoleted a long time ago by '2 pass Conditional Bach'?

I have been trying to be more careful lately and use the term OPV Q-Value when referring to OPV mode Q. and Q-Factor when referring to the bitrate analysis tab in CCE.

I think also, if you asked 3 people what RoBa is, you will get three different answers. The funny thing is, they could all be right in a way because of the way it morphed and evolved as time went on. That's some of us have tried to quit using the name as it no longer differentiates adequately.

[influenza]

I have been trying to be more careful lately and use the term OPV Q-Value when referring to OPV mode Q. and Q-Factor when referring to the bitrate analysis tab in CCE.

I should remember this. Exactly what i meant ;)

[Eyes`Only]

Well you'll never hear me talking about OPV anything, as I'm not a fan of it at all. But just for the sake of it, you suggest Q-Factor? Why not something that's more to the point like Robshot-Q?

Or if you want to get all technical.. CCE itself calls it "Q. Scale".

Anyway, yes I meant the Q-Scale/Q-Factor/Robshot-Q ;)

[DDogg]

Well you'll never hear me talking about OPV anything, as I'm not a fan of it at all. Sure would like to hear your technical rationale for that statement. I'm always eager to learn something.

[Lagoon]

I don't like the Q9 rules from robshot at all. There are LOTS of mosquito noise and some blocks here and there easily viewable on my TV.

Anyway, the minimum bitrate I used was 1600, for a 2.35:1 movie with lots of dark scenes, had a Q of 5 even with that low bitrate.

An acceptable bitrate for a one-movie disc is 3000 for me, that can get a 2h30 movie on fine.

[Trahald]

we are referring to robshot-bach method ( http://forum.doom9.org/showthread.php?s=&threadid=57518&highlight=roba ) <-interesting read explains its origons. it may be a little hard to follow since bach removed all of his replies but thankfully most are quoted back in

basically its saying (amongst other things) to allow cce to decide what avg bitrate (and ultimately size) is good for you with a opv pass based on the quality you want not the size. ultimately in a dvd environment you are (sortof) stuck size wise. if you choose a (opv) q-value of 30 or better(ie lower) acceptable quality, and encode your file with that, and find out the size is 1000 mb more than the space you can realistically give your movie (meaning youve gotten rid of every foreign language track and lowered the extras bitrate to as low as you can stand(wether that be opvqvalue-wise or bitrate-wise) and you still cant meet the goal.. you are either forced to use a second disk (which is acceptable for svcd already being a multiple disk environment but generally not dvd) or removing extras / doing the main movie at 1/2 d1 / maybe adding filtering.. or some combination. these options may or may not be acceptable to you..

bach also felt the opv pass produced a very optimized output. an additional pass may help tweak bitrate at gop level but any more would actually add distortion

over time it has become many things. as ddogg says if you ask 3 people what roba is you will get 3 different answers

i use roba + 1pass on all my encodes and i think the output is great.

[DDogg]

If you like the many benefits of RoBa/Bach, but also like the perfect size accuracy of multipass, you may find some additional information in the link in my sig that can help you combine both in a synergistic way. Very similar to 2pass Bach, but with a few slight twists that allow a more simple and direct method using multipass 1pass (VAF+1 pass) to achieve a similar result, quality and time wise.

[Eyes`Only]

Originally posted by DDogg
Sure would like to hear your technical rationale for that statement. I'm always eager to learn something.

Simply put, I find continuously better results using multiple passes on my extras at 1750-1850 ABR. Usually 5. Tried the oft-heralded one-pass method on my extras and it looked like DDoggsh**. (Sorry, I couldn't resist) :)

As an extra note... I actually found better results using multipass vs. Roba for them too. Go figure.

[DDogg]

Man, that was a real technical rationale, my head is spinning from all those heavy implications :) Speaking of stinky stuff...:D

[dazzle]

Since I am a projector user (which is nice, but it also shows all the flaws in 120":)) I never go below 8 passes. I also try to keep the avg bitrate around or above 4000 if possible, depending on a lot of things of course (length, extras, menu etc).

If bitrate tends to be low, I always rip the extra stuff, unnecessary audio etc. This since I care for the movie, and it's presentation, and not often of the bonus stuff. When I do keep extras I usually use 5-6 passes.

Works for me! :)