BD Rebuilder Beta - Bug Reports Only

[Chuckwagon]

Originally Posted by jdobbs

When are you making the manual selections? If you do it after some encoding have been completed and you resume, I can see how it would be off because the size of the completed items wouldn't match the new calculations.

Any chance that's it? It sounds like it -- and if that's the case it make perfect sense, the only thing I could do is start the entire encode over when you change your selections. You can't change the size of a file that has already been encoded...

To test this I shutdown BD-RB, and I make sure the target directories are empty. I mount the disc to be processed. I start BD-RB, it's running in Movie and Menus Backup mode, and it scans the disc and selects the streams it thinks it should keep. I double click the title to put it into edit mode, and then I do the manual selections and hit backup. No pre-processing takes place, no interruptions or resumptions of the rebuild process happen. No matter if I add or delete streams, whatever input size BD-RB calculated when it started and did its auto blanking is the size it chooses when I start the backup.

The source for the x264 tests was a single file on the SSD array. The source I use for my rebuilds are ISO files made with Anydvd that reside on a regular HDD.

[jdobbs]

Originally Posted by Chuckwagon

To test this I shutdown BD-RB, and I make sure the target directories are empty. I mount the disc to be processed. I start BD-RB, it's running in Movie and Menus Backup mode, and it scans the disc and selects the streams it thinks it should keep. I double click the title to put it into edit mode, and then I do the manual selections and hit backup. No pre-processing takes place, no interruptions or resumptions of the rebuild process happen. No matter if I add or delete streams, whatever input size BD-RB calculated when it started and did its auto blanking is the size it chooses when I start the backup.

The source for the x264 tests was a single file on the SSD array. The source I use for my rebuilds are ISO files made with Anydvd that reside on a regular HDD.

I'll check it out.

[jdobbs]

Originally Posted by Chuckwagon

To test this I shutdown BD-RB, and I make sure the target directories are empty. I mount the disc to be processed. I start BD-RB, it's running in Movie and Menus Backup mode, and it scans the disc and selects the streams it thinks it should keep. I double click the title to put it into edit mode, and then I do the manual selections and hit backup. No pre-processing takes place, no interruptions or resumptions of the rebuild process happen. No matter if I add or delete streams, whatever input size BD-RB calculated when it started and did its auto blanking is the size it chooses when I start the backup.

The source for the x264 tests was a single file on the SSD array. The source I use for my rebuilds are ISO files made with Anydvd that reside on a regular HDD.

Thanks for this feedback. After I read your post I started wondering how that could be possible and traced the code... arghhh -- it was so obvious I don't know how I could have missed it before. I found just about the silliest error of all time. When you select "blank/unblank" a routine finds all playlist links and changes their flags. The sizing numbers are then adjusted for all the linked streams. BUT THE ROUTINE NEVER ADJUSTS THE SIZES FOR THE ONE YOU ACTUALLY SELECTED!!!

Sometimes I just have to laugh at myself.

I'll fix it for the next release.

Thanks again for the report.

[Sharc]

Currently I get a problem with CRF prediction for a multi-segment movie. For one of the segments the prediction converges in one single step from 22.48 (initial value) to the final value of 36.24 and the quality is accoridingly poor. Therefore I try to get around it by adjusting parameters.
According to hiddenoptions.txt the 2 params are:
GROUP_SIZE=
SAMPLE_SIZE=
According to the original post here the 2 params are:
SAMPLE_GROUP=
SAMPLE_SIZE=
Which syntax is the correct one?

Edit:
I understood that for "very short" segments BD-RB will fall back automatically to 2-pass. However it seems that it does a CRF prediction for short segments as well. Is there a parameter which defines the size of a "short segment" below which a 2-pass encode is automatically done in place of the CRF prediction?

[jdobbs]

Originally Posted by Sharc

Currently I get a problem with CRF prediction for a multi-segment movie. For one of the segments the prediction converges in one single step from 22.48 (initial value) to the final value of 36.24 and the quality is accoridingly poor. Therefore I try to get around it by adjusting parameters.
According to hiddenoptions.txt the 2 params are:
GROUP_SIZE=
SAMPLE_SIZE=
According to the original post here the 2 params are:
SAMPLE_GROUP=
SAMPLE_SIZE=
Which syntax is the correct one?

Edit:
I understood that for "very short" segments BD-RB will fall back automatically to 2-pass. However it seems that it does a CRF prediction for short segments as well. Is there a parameter which defines the size of a "short segment" below which a 2-pass encode is automatically done in place of the CRF prediction?

It should be SAMPLE_GROUP and SAMPLE_SIZE. SAMPLE_SIZE is the size of the sample to be used and SAMPLE_GROUP is how bit a segment each sample represents. Not sure how/when GROUP_SIZE got thrown into the HIDDENOPTS.TXT file -- but it doesn't exist. I'll fix that.

The default SAMPLE_SIZE is 2x the source framerate and the default SAMPLE_GROUP is 200x the source framerate. So a 24fps source would be sampled every 4800 frames using a sample of 48 frames. If you set a fixed sample group/rate, however -- it uses what you've set regardless of the input frame rate.

[Sharc]

Originally Posted by jdobbs

It should be SAMPLE_GROUP and SAMPLE_SIZE. SAMPLE_SIZE is the size of the sample to be used and SAMPLE_GROUP is how bit a segment each sample represents. Not sure how/when GROUP_SIZE got thrown into the HIDDENOPTS.TXT file -- but it doesn't exist. I'll fix that.

The default SAMPLE_SIZE is 2x the source framerate and the default SAMPLE_GROUP is 200x the source framerate. So a 24fps source would be sampled every 4800 frames using a sample of 48 frames. If you set a fixed sample group/rate, however -- it uses what you've set regardless of the input frame rate.

ok, thanks for explaining.
This means that by default any segment shorter than 4800 frames (for a 24 fps source) would be based on the first 48 frames for CRF prediction, correct? If these first 48 frames are of simple structure (e.g. fading from black) we could end up in very biased CRF prediction for this segment?
Perhaps segments of <4800 frames should be automatically 2-pass encoded, or 4800 should be auto-reduced such that at least 2 groups of 48 frames are used for CRF prediction?
Normally CRF prediction works very well , with occasional exceptions for segmented sources (small segments). Perhaps too complicated ....

[Chuckwagon]

Originally Posted by jdobbs

arghhh -- it was so obvious I don't know how I could have missed it before.

Heh heh, it's because the obvious ones we just correct mentally as we sail past them, they are so obvious after all that they can't really be wrong.

Glad you were able to find it, and happy to be of help. And thank you for such an awesome tool.

[DarWun]

FYI, as a test I just did a full BD25 backup of Let Me In using BD Rebuilder 39.03. Since it has PIP with DTS-Express secondary audio on the main feature, I set DTSX_ENABLE to 1. The burned disk worked flawlessly on my PS3.

[jdobbs]

Originally Posted by Sharc

ok, thanks for explaining.
This means that by default any segment shorter than 4800 frames (for a 24 fps source) would be based on the first 48 frames for CRF prediction, correct? If these first 48 frames are of simple structure (e.g. fading from black) we could end up in very biased CRF prediction for this segment?
Perhaps segments of <4800 frames should be automatically 2-pass encoded, or 4800 should be auto-reduced such that at least 2 groups of 48 frames are used for CRF prediction?
Normally CRF prediction works very well , with occasional exceptions for segmented sources (small segments). Perhaps too complicated ....

No. Since the beginning and the end are usually biased, they are not used. So you'd end up with 48 frames from the middle somewhere. It's really not very often that you find a stream that is only 3 minutes long anyway. On a more typical movie (say 2 hours) you're going to be sampling 36 sets of 48 frames that are spread throughout the stream, so it's unlikely any of them would carrry enough weight to really skew it.

The default does a 1% sample -- but you could always make it 2% or even 3% (e.g. SAMPLE_GROUP=1200, SAMPLE_SIZE=48). Be careful with smaller sample sizes, you really don't want to get below GOP sizes as it can make prediction less accurate.

[Sharc]

Originally Posted by jdobbs

No. Since the beginning and the end are usually biased, they are not used. So you'd end up with 48 frames from the middle somewhere. It's really not very often that you find a stream that is only 3 minutes long anyway. On a more typical movie (say 2 hours) you're going to be sampling 36 sets of 48 frames that are spread throughout the stream, so it's unlikely any of them would carrry enough weight to really skew it.

Agree for the more typical movie. I came however across the problem with segmented movies (i.e. theatre and extended version controlled by playlist) where some segments are shorter than 4800 and lead to very skewed CRF predictions between these segments (ranging from CRF=10 to 36). The larger segments (>4800 frames) came out as expected with a CRF around 19 (for DVD-5/720p). Therfore I thought that short segments should automatically be encoded 2-pass.
Never mind, such cases should perhaps better be done using 2-pass a priori rather than 1-pass CRF predictive mode (I just like it and tried.... ).

[Stereodude]

Originally Posted by Sharc

Never mind, such cases should perhaps better be done using 2-pass a priori rather than 1-pass CRF predictive mode (I just like it and tried.... ).

Isn't using 2-pass encoding on a movie in multiple pieces a bad idea since each piece has no visibility to any other piece and some parts could end up wasting bitrate that would be better used elsewhere netting inconsistent visual quality between the pieces? It seems like CRF would be the best way to maintain quality between the various pieces with the challenge being the determination of what CRF value to use (across all the pieces).

[jdobbs]

Originally Posted by Stereodude

Isn't using 2-pass encoding on a movie in multiple pieces a bad idea since each piece has no visibility to any other piece and some parts could end up wasting bitrate that would be better used elsewhere netting inconsistent visual quality between the pieces? It seems like CRF would be the best way to maintain quality between the various pieces with the challenge being the determination of what CRF value to use (across all the pieces).

Not really -- mainly because you have the original to use as a baseline (that has already been weighed and encoded). Also -- it doesn't take a whole lot of frames to benefit from 2 pass.

[santito]

hi...

i can confirm that the blurays are stopped to work in the latest versions in some players... i have a lg bd-370 and the titles (not all.. 90%) get black screen...
but in my ps3 works fine...

in the tv... i get a black screen.. and in the screen of player the counter is in 0:00:00:00

i will try to encode with 38.08... i will report tomorrow..

[jdobbs]

Originally Posted by santito

hi...

i can confirm that the blurays are stopped to work in the latest versions in some players... i have a lg bd-370 and the titles (not all.. 90%) get black screen...
but in my ps3 works fine...

in the tv... i get a black screen.. and in the screen of player the counter is in 0:00:00:00

i will try to encode with 38.08... i will report tomorrow..

Confirm? What report are you confirming? What kind of backups? Movie-only, full? Backups to BD-25, BD-9, BD-5? Do you have "Strict AVCHD" set? Are you encoding from originals? What does "latest versions" mean -- can't you just say the version you're using? You'll "try 38.08"? Why?

This is a thread for reporting bugs... with necessary information to make fixes. How can I fix something when I have no idea what it is.

[santito]

hi...

sorry i was not very specific...

i use 39.03 ....full movie backup (bd-25)...

i guess is not my encode machine (pc).. because others have reported the same issue... not working in that particular player (lg bd-370) and panasonic player

[jdobbs]

Originally Posted by santito

hi...

sorry i was not very specific...

i use 39.03 ....full movie backup (bd-25)...

i guess is not my encode machine (pc).. because others have reported the same issue... not working in that particular player (lg bd-370) and panasonic player

I'm not sure where you're seeing the "others" -- because this is the bug thread and I don't see it here. It sounds like your player doesn't support the type of disc you're using. What make/brand is it? If using Imgburn, report what it says about the empty disc.

[drmih]

I have a LG bd-370 which I don't often use, but I have just checked a bd-25 disc I did with 39.03 and it was okay. However, the LG is a bit choosy and certainly on bd-9 only plays discs without jar files. It may be that it has problems with certain discs.

[wooper]

Hi there, i have a question , hope i dont bother anyone with this, is just a simpel question, how can i set bdrebuilder to work using dual-procesor, i mean, using 64 bit operating system, i found that in tool folder you have x264-64 and setacl_x64, is it a way to set those so i can rebuild movies faster? just a question, thanks

[Atak_Snajpera]

setacl is not an encoder. it is only used for setting ffdshow as preffered decoder in windows 7.

[jwjohnson]

You don't have to set anything. x264 uses all the cores available.