Will BD Rebuilder support MPEG2?

[rendez2k]

I tested another new tool yesterday that takes a BR structure and converts it to BR25/9/5 (similar to BD Rebuilder) but it uses MPEG2 rather than x264. Now I know the output probably isn't as good (but it looked great to me) but it only took around 3 hours to do a 2-pass encode. I see FFMPEG Encoder is greyed out so I assume it will be added?

[pbeumer2001]

I tested another new tool yesterday that takes a BR structure and converts it to BR25/9/5 (similar to BD Rebuilder) but it uses MPEG2 rather than x264. Now I know the output probably isn't as good (but it looked great to me) but it only took around 3 hours to do a 2-pass encode. I see FFMPEG Encoder is greyed out so I assume it will be added?

What prog did you use?

[shon3i]

Why MPEG2? i mean it wrost possible encoder, and i can imagine 1080p video on BD-5 in MPEG2:)?? BDRebulder is early beta stage, and now support only H264/AVC, but probably will later have ability to keep original streams untoched.

btw what app you test, share with us :)??

[jdobbs]

Supporting MPEG-2 is an easy thing to do. Remember though, that even the outstanding encoders like HC Encoder can't compete with AVC for quality at the same bitrate. For example, I did a movie-only backup as a test last night and the overall bitrate was 4700Kbs. That would be great for a DVD @ 720x480... but could you imagine how it would look @ 1920x1080p? That's 2 million pixels to be encoded for each frame as compared to 346 thousand.

Just for the heck of it, I'm doing that same movie with HC as a test. From what I see so far it looks like it is going to take about five hours. The X264 encode took eight on the same machine.

I plan to support HC Encoder from within BD Rebuilder at some point. The reason is mainly for BD-25 backups.

[rendez2k]

It was a BD25 @ 1920x1080 movie only backup I did. I can imagine it would be horrible on BD-5 though!

pbeumer2001 + shon3i: http://www.videohelp.com/tools/D2MP

[Dark Shikari]

I tested another new tool yesterday that takes a BR structure and converts it to BR25/9/5 (similar to BD Rebuilder) but it uses MPEG2 rather than x264. Now I know the output probably isn't as good (but it looked great to me) but it only took around 3 hours to do a 2-pass encode. I see FFMPEG Encoder is greyed out so I assume it will be added?If you want faster encoding, I'm pretty sure that x264 can outperform most MPEG-2 encoders if you give it fast enough settings.

The simple fact that H.264 uses a transform that's about 3x faster should ensure that MPEG-2 never wins in a speed test. I don't see speed as a valid reason to support MPEG-2 in BD-Rebuilder (I can imagine other reasons though).

[jdobbs]

I think a lot of people confuse the speed issue because they use X264 for HD and MPEG-2 for SD, and forget the difference in resolution. I can tell you that X264 is very, very fast on SD encodes. In fact, I'll probably add it to DVD Rebuilder so you can make much more "perfect" movie-only backups of DVDs to AVCHD or BD-5.

My own experience is that there is no MPEG-2 encoder that even comes close to X264 in terms of quality (even the better products like CCE, MainConcept, or HC) at comparable bitrates.

[rendez2k]

It wasn't really the speed that I was interested in - I'm all for spending time to get better results. I just wanted to test the software and results and on the BD25 test I did, it actually looked a lot better than I was expecting! So I wondered if it was a viable alternative way of encoding BD25s.....

[jdobbs]

I just stopped the MPEG-2 encode. It would have taken a little over 4 hours. But the output is very blocky... 1080p at 4700Kbs isn't going to work with MPEG-2. But there are a lot of commercial discs out there that use it for BD-50 encodes -- so it may be ok for BD-25. It won't compare to AVC, though.

[rendez2k]

I just stopped the MPEG-2 encode. It would have taken a little over 4 hours. But the output is very blocky... 1080p at 4700Kbs isn't going to work with MPEG-2. But there are a lot of commercial discs out there that use it for BD-50 encodes -- so it may be ok for BD-25. It won't compare to AVC, though.

Was that to BD5?

[shon3i]

Was that to BD5?
BD25 or BD5, there is no point to use MPEG2 because AVC is always better on all bitrates, and AVC with normal settings will give you speed and quality, when mpeg2 can give yu little more speed but quilaty is worse

[Furiousflea]

I really don't see the point in using MPEG2 encoding ever regardless of the speed increase. If you change the x264 settings down low enough you will get comparible speed and still have better picture quality. Am I missing something?

[~bT~]

this is such a backward thread. mpeg2?? "you havin' a laff?"

anyway, pls add avchd to dvd-rb jdobbs. cheers!
hope you add smaller size options apart from dvd5/9.

[asarian]

I really don't see the point in using MPEG2 encoding ever regardless of the speed increase.

Can't exactly say I see the use of MPEG2 encoding, either (apart from real-time transcoding for streaming media).

[d2mp]

It was a BD25 @ 1920x1080 movie only backup I did. I can imagine it would be horrible on BD-5 though!

pbeumer2001 + shon3i: http://www.videohelp.com/tools/D2MP

In the next days a new version of D2mp will be released with an option to choose between H264 and MPEG2 outputs.
But I can say H264 is muuuuuccccchhhhh sloooowwwwweeeerrr when compared to MPEG2 and not so much diference.

;)

(Tested with a PS3, and a 42" Panasonic Plasma TV)

[Dark Shikari]

In the next days a new version of D2mp will be released with an option to choose between H264 and MPEG2 outputs.
But I can say H264 is muuuuuccccchhhhh sloooowwwwweeeerrrI just did a test and x264 was only about 10-15% slower than libavcodec MPEG-2, both on max speed settings...

I wouldn't call that "muuuuuccccchhhhh sloooowwwwweeeerrr", maybe "somewhat slower"?

[d2mp]

I just did a test and x264 was only about 10-15% slower than libavcodec MPEG-2, both on max speed settings...

I wouldn't call that "muuuuuccccchhhhh sloooowwwwweeeerrr", maybe "somewhat slower"?

Well, maybe I'm doing something. and if so, I will apreciate any inputs:

For the same source (a 1920x1080 MPEG4/ISO/AVC)
With the same encoding tool (ffmpeg).
Using a Dual Core T5600 IntelCPU.

MPEG2:

FFMPEG.EXE -i SOURCE.AVS -y -an -threads 2 -b 15000k -maxrate 20000k -bufsize 4096k -pass 1 -vcodec mpeg2video -dc 8 -aspect 16:9 -f rawvideo NUL

getting 10 fps and 67% CPU occupancy.


libx264:

FFMPEG.EXE -i SOURCE.AVS -y -an -threads 2 -b 15000k -maxrate 20000k -bufsize 4096k -pass 1 -vcodec libx264 -flags2 dct8x8 -aspect 16:9 -f rawvideo NUL

getting 5 fps and 100% CPU occupancy.
So, Half the speed.


Thanks a lot

[Dark Shikari]

Well, maybe I'm doing something. and if so, I will apreciate any inputs:

For the same source (a 1920x1080 MPEG4/ISO/AVC)
With the same encoding tool (ffmpeg).
Using a Dual Core T5600 IntelCPU.

MPEG2:

FFMPEG.EXE -i SOURCE.AVS -y -an -threads 2 -b 15000k -maxrate 20000k -bufsize 4096k -pass 1 -vcodec mpeg2video -dc 8 -aspect 16:9 -f rawvideo NUL

getting 10 fps and 67% CPU occupancy.


libx264:

FFMPEG.EXE -i SOURCE.AVS -y -an -threads 2 -b 15000k -maxrate 20000k -bufsize 4096k -pass 1 -vcodec libx264 -flags2 dct8x8 -aspect 16:9 -f rawvideo NUL

getting 5 fps and 100% CPU occupancy.
So, Half the speed.


Thanks a lotSo to compare to x264, you proceed to:

1. Use a terrible frontend with awful defaults (and do nothing to resolve this, like specifying them correctly). This hurts x264 in your test.
2. Encode from Avisynth input, thus testing a combination of encoding and decoding, rather than just encoding. This probably hurts mpeg-2 in your test.
3. Use a completely unknown and unspecified version of ffmpeg and x264. This probably hurts x264, because with the odds these days, you're probably using a 2 year old version of x264, and ffmpeg hasn't had any significant changes to the core mpeg encoder in that time.
4. Use the wrong number of threads (use auto).
5. Use a really weird bufsize (dunno how this affects things, probably just borks quality.)

My test was on Gentoo 64-bit and used "--no-cabac --nf --partitions none --me dia --merange 8 --subme 0 --qp 24" for x264 and default settings for ffmpeg's mpeg encoder (as, afaik, defaults are basically chosen to be fastest), along with -qscale 3 for the mpeg encoder, as that is roughly the MPEG qp that corresponds with qp 24 in H.264. I did the test without threads.

I'll do another test with target bitrate mode and threads, like you did.

[Dark Shikari]

Test on http://imk.cx/videos/wic_1280x720_30.264 decoded into raw video (y4m) before encoding. Thanks to Dopefish for the clip--I didn't have any good HD raw clips lying around on a server I could easily wget from, and my upload sucks.

Latest ffmpeg and x264:

ffmpeg VBV+bitrate15k: 127fps (used full CPU power because mine was properly compiled with threads support ;) )
ffmpeg qscale 3: 89fps (bitrate 57.5k)
x264 VBV+bitrate15k: 86.5fps
x264 qscale 24: 105fps (bitrate 56.5k)

It'd be a better test with a better input clip though--that one is way too noisy, leading to the really high bitrates.

Conclusion:

1. ffmpeg's threading for mpeg-2 encoding works pretty damn well.
2. x264 loses loads of speed going into bitrate/VBV mode. ffmpeg actually gains speed due to the lower relative bitrate. I'm going to guess ffmpeg's lookahead is far faster than x264's--which it is, because what ffmpeg does is that it makes lookahead and the "regular motion search" one and the same. So it costs nothing to do the lookahead necessary for bitrate mode encoding.

[Sharc]

I think d2mp can be very useful for re-authoring BluRay movies to standard DVDs at 720x480 (576) resolution which can be played on any legacy DVD standalone player.

[jdobbs]

Was that to BD5?Yes, it was a BD-5. I do that when the bitrate is decent. 4700Kbs looks good for 1080p when used with AVC.

[Sharc]

I just stopped the MPEG-2 encode. It would have taken a little over 4 hours. But the output is very blocky... 1080p at 4700Kbs isn't going to work with MPEG-2. But there are a lot of commercial discs out there that use it for BD-50 encodes -- so it may be ok for BD-25. It won't compare to AVC, though.
I resized a BD 1920x1080 (VC-1) to DVD 720x576 resolution using spline36resize, and encoded to mpeg2@3500 kbps with HCenc. I am absolutely surprised about the fine quality.
(Intention is to create a backup of BDs which I can play on my legacy DVD standalone).

[Furiousflea]

I resized a BD 1920x1080 (VC-1) to DVD 720x576 resolution using spline36resize, and encoded to mpeg2@3500 kbps with HCenc. I am absolutely surprised about the fine quality.
(Intention is to create a backup of BDs which I can play on my legacy DVD standalone).

Of course it would be. The quality of the BD disc as a source is far superior in general to the tosh that gets used for making DVD movies. I'm sure that recent DVDs released in past year might be different story...It's only with the release of HD-DVD\Blu-Ray that studios have spend any real money on restoration so your source is going to be way better than theirs. Also, they always pump up their release with far too many extras\audio tracks and junk so your 3.5mbit bit rate isn't actually that low compared to the retail DVD release. Finally the quality of some (most) DVDs is abysmal makes you think that there must have been a troop of monkeys at the machine tweaking the encode for the DVD release on many occassions.

[Sharc]

Agreed!

[~bT~]

^ with dvd's, i was using a few filters to improve pq here n there but with bd, i don't use any filters.