Which bitrate for HD material?

[Devilman1]

I'm wondering how to calculate the best bitrate to encode 1080p blu ray stream without resize it?

Thanks

[Atak_Snajpera]

Why do you want to calculate 'best bitrate' if there is CQ mode? Start with CRF@22. If you really need 'best bitrate' I suggest to not go below 12Mbps. If movie has wider aspect ratio (2.35:1) use 8Mbps.

[Devilman1]

I didn't understand much of x264 option like cq mode. I'm still linked to old enconding way with bitrate.

[Egladil]

Then read the documentation and the forums to educate yourself about the x264 options. Bitrate-based encoding is dying for HD stuff, because the bitrate totally depends on the source. sharp, crispy and fast require way more bitrate than soft and slow movies. The variation can be really huge.

Use megui or other x264 gui and try the constant quality mode, as Atak_Snajpera said, but with crf 22 your encode will look like shit, use crf 18 for an almost transparent encode.

[Atak_Snajpera]

but with crf 22 your encode will look like shit, use crf 18 for an almost transparent encode.

not really if you combine CRF 22 with AQ (0.5 ; 5) . Size will be very similar to 18 but you won't see ugly blocks in dark/blue areas.

[Chainmax]

If he needs to reach a certain filesize, then two-pass encode is still needed.


Devilman1, bitrate depends on the source type (animation, real-life footage) as well as the amount of motion in it, the detail in the source and many, many other factors. It also heavily depends on how/if you filter the sourceand what x264 options you select. Hence, there isn't really a best overall bitrate, especially taking into account that "best" is subjective.

Still, some generalized recommendations might be made. I'd suggest you to start at 0.16b/pf and see how it looks, then adjust bitrate accordingly. What does that mean? Well, for a 720p source @23.976fps you'd have:


(bitrate * 1024) / (Hres * Vres * fps) = b/pf

(bitrate * 1024) / (1280*720*23.976) = 0.16

bitrate*1024 = 0.16*1280*720*23.976

bitrate = (0.16*1280*720*23.976)/1024

bitrate = ~3450kbps

[Sagekilla]

Personally, I find bpp to be a very bad measure to determine bitrate or quality because of how varied your results can be depending on your source. I really don't see a point in using it, since if you're aiming for a specific size you wouldn't need it. On the flip side, if you're not aiming for a specific size you can just use crf for constant quality regardless of the source video's complexity. Using crf, you can just encode within the range of crf 18 - 22, with 18 being pretty much transparent and 22 being quite less so.

Edit: Chainmax, I believe there was a post before that provided an explanation as to why bpp wasn't a good measure and you should use a weighted one instead. I forgot where it was, but I'll post it here when I do.

[ToS_Maverick]

i personally use abr 1pass with
selectrangeevery(1200,12)
in the script as quality check. in the log you'll get the final ratefactor. i try to aim at ~20.

[Atak_Snajpera]

i try to aim at ~20.
Then Why don't set CRF@20 instead of abr?????

[ToS_Maverick]

because i aim at DVD5. i need the check to choose the resolution. lower one for longer or hard to compress movies, higher res for easy to encode movies.

[Devilman1]

Thanks for all suggestion.
I still use the bitrate with 2-pass because often I need file not bigger than a specified size and with abr or crf, if I understand well, this is not possible.

thanks Chainmax for the formula.

[cogman]

you'll get much better results with a two pass encode. Yes it takes longer, but you should see a better result overall.

[Dark Shikari]

you'll get much better results with a two pass encode. Yes it takes longer, but you should see a better result overall.2-pass is not at all better than CRF at the same bitrate. It is much better than 1-pass ABR though.

[Sagekilla]

2-pass is not at all better than CRF at the same bitrate. It is much better than 1-pass ABR though.

Could be argued crf is spending the bits better too, since you can get away with allocating fewer bits to really fast sequences and have it still look good (which is what crf does) and then allocate just enough bits to make a lower motion scene look good while not spending excessive amounts.

[cogman]

opps, I guess I should have been more clear on who I was directing my comments to. ToS_Maverick was who I was aiming at as I just can understand why you would use ABR (unless you are streaming the video over the internet or something, then it makes a little sense)

[ToS_Maverick]

i use ABR only for my analyse pass, which only encodes 1% of the movie. after i got my ratefactor, i probably resize (if to high), or start the 2pass encode.

sry if i wasn't clear enough

[Sharktooth]

why not just use CRF for analysis pass?
1-pass ABR is just the worst mode...

[ToS_Maverick]

because abr is very exact in pointing out the ratefactor! i only need 1 try to get the number, where i have to trial-and-error with CRF.

please see this thread, where DDogg did a great research on prediction with x264:
http://forum.doom9.org/showthread.php?t=116773

[gwaitsi]

maybe some one could answer for me please.

I use nero recode and found (for normal DVD) 1.5mb/s 2pass to provide the best result with about 1.6Gb for 1.5hr movie at encoder precision good, or 1.2Gb for the same movie with encoder precision best.

if i want to use this crf instead of 2pass, can i make nero recode encode in the way, and how to switch to crf encoding please?

[Inventive Software]

CRF is x264-specific AFAIK. I've found CRF 20 to be more than adequate for most movies, this gives roughly an average bitrate of 1.5 - 2 Mbit/sec.

[Razorholt]

More people will use CRF (vs 2-pass) the day someone will find out a way to predict the final size of the encoded file.

I will...

[Sharktooth]

because abr is very exact in pointing out the ratefactor! i only need 1 try to get the number, where i have to trial-and-error with CRF.

please see this thread, where DDogg did a great research on prediction with x264:
http://forum.doom9.org/showthread.php?t=116773
i still cant understand. a CRF analysis pass is still better than ABR.
the final target is to encode at a desired quality and see if we can fit the movie in a certain media.
good, then a compression test (or analysis pass... or whatever) can be done on a small % of the move at the desired CRF to predict the final filesize and see if it will fit in your media.
if it doesnt, you can rise the CRF value to an acceptable level, filter the source, reduce the resolution, ecc...

[Inventive Software]

Question about CRF: what's the maximum deviation the QPs are from the target CRF number?

[ToS_Maverick]

@Sharktooth
are you suggesting that i should encode my movies in CRF mode? because normally i use the Ratefactor as indicator, and encode in 2pass after the comp. check.

[Dark Shikari]

@Sharktooth
are you suggesting that i should encode my movies in CRF mode? because normally i use the Ratefactor as indicator, and encode in 2pass after the comp. check.Why do people not understand this? Its simple as this:

1. If you want a specific quality, encode in CRF.

2. If you want a specific filesize, encode using twopass ABR.

The subjective/objective quality difference between a CRF and a 2pass ABR at the same bitrate is negligable. The subjective/objective quality difference between a 1pass ABR firstpass and a CRF firstpass is negligable. Just do what makes the most sense for you.

[Sharktooth]

im suggesting to always run comp check for bitrate modes and encode in CRF mode whenever is possible.

[Dark Shikari]

im suggesting to run a comp check and then encode in CRF mode whenever is possible.Why? That's really not much different from running a 2pass ABR, except the result is slightly less exacting.

[Sharktooth]

reworded the previous post to be more clear.

[ToS_Maverick]

thank you for the clear explanation!

since i want to use most of the DVD size, i encode in 2pass

[Sagekilla]

For the record, I prefer CRF because:

1) I get quality, guaranteed
2) It's faster than 2-pass since crf is practically same speed as the 2nd "final encode" pass.
3) When you're dealing with pre-processing on avisynth, crf wins by a HUGE margin when it comes to the really slow scripts.


Edit: Hope that helps anyone trying to make a decision, but you really shouldn't use crf unless you're aiming for a specific size. Unless you have a good understanding of the expected bitrate (this can vary wildly, however) you shouldn't use CRF when fitting for DVD.

[Egladil]

...
3) When you're dealing with pre-processing on avisynth, crf wins by a HUGE margin when it comes to the really slow scripts.
...


For slow scripts I encode to HUFFYV12 and then use this for x264. For 720p the HUFFYV12 file is rarely over 150 GiB

[StephenChow]

I tested some HD clips which are 720p, and I believe that 1.500 Kb/s-2.000 Kb/s is OK for 720p

[mitsubishi]

I tested some HD clips which are 720p, and I believe that 1.500 Kb/s-2.000 Kb/s is OK for 720p

Not really, maybe cartoons or something (I don't do anime etc), but for most material I've encoded 3-6mb/s is a more normal range for decent quality. Even "the scene" rules minimum bitrate for 720p TV works out about 2.5mb/s (4 for films)