BD Ripper (3D Plus) Version 9028 - h.264 mp4 encodes use only half of the target bitrate.

[lalittle]

I've noticed in my tests with DVDFab 9028 that mp4 encodes using the h.264 encoder are not obeying the bitrate/filesize values that are used in the DVDFab advanced encode settings. I would expect to see "some" varience from the target values, but the results I'm seeing are WAY off, with the resulting files having average bitrates and file sizes that are about HALF -- or even less -- of the values that were entered.

I did not see this issue with version 8228 -- the resulting bitrates and file sizes for h.264 encodes were very close to the values entered in the DVDFab advanced encode settings. Unfortunately, 8228 has other issues with h.264 encodes (particularly with mp4 encodes), that are fixed with 9028, so I'm forced to use version 9028. (Details of this are in another thread.)

I check the bitrates and other info with both MediaInfo and Bitrate Viewer, and these programs agree on the results. I tried more than one movie as well, and saw very similar results. For example, using a setting of 15,000 resulted in an average bitrae of only 8,850, while on another movie a setting of 20,000 gave me only 9,445 Kbps -- less than half of the target value.

On a related note, version 9028 is using the "Main" h.264 profile instead of the much preferred "High" profile that was used in 8228. If only one profile is offered, it should be the higher quality "high" profile.

I'm using 2-pass for all my tests, and have disabled all hardware acceleration for both decoding and encoding.

Larry

[90312]

Larry, I've tested this a few times today and it has worked pretty good. I set the bit rate to 20,000 and it comes out 19,400 or so. Are you setting like in the snap below:

[lalittle]

Thanks for the reply.

The only difference I can see is that I'm using 2-pass, which you used 1-pass. This could be the difference.

I'll do some tests to see if using 1-pass results in correct bitrates for me, but that wouldn't solve the issue because 2-pass produces superior results.

In the mean time, do you happen to know if your result is using the Main profile instead of the High profile? As I mentioned above, 9028 is using the main profile, whereas 8228 used the high profile (higher quality) in the same situation.

Thanks,

Larry

[90312]

At the bit rate you're using Larry you'll never see the difference between 1 and 2 pass encoding.

[lalittle]

At the bit rate you're using Larry you'll never see the difference between 1 and 2 pass encoding.

Nevertheless, it should work with 2-pass.

On that note, 2-pass always has the potential to look better, and in my situation, I need to make the effort to get the best results I can.

Thanks,

Larry

[lalittle]

I just did a test using 1-pass instead of 2-pass, and my result was much closer to the target. Using a setting of 15,000Kbps, I ended up with a file with an average bitrate of 12,900Kbps, which is within normal limits.

My next test, however, showed some surprising results. I used the same chapter of this BD using 2-pass with the rest of the settings the same. Using the same target bitrate of 15,000, it resulted in a file with an average bitrate of 13,900 -- i.e. a normal result rather than only half of the target bitrate that I saw with the previous disc.

I then checked the profiles, and both of the "new" rips used the high profile rather than the main. After several tests, I'm finding that some end up with high profiles, while others use the main profile. I'm using all the same settings, so I don't understand why this is happening. DVDFab is using different profiles for different rips, and I can't see any pattern to why it uses one vs the other.

The only difference between the tests was the BD itself (my earlier test resulting in low average bitrates was a different movie.) I now need to go back and do another test of the previous disc and see what happens. I don't know why this would make a difference, but one movie gives me results of only half the target bitrate, while the other movie stays on target. Whether I end up with the main or the high profile seems to be random. Both discs are VC-1 encodes.

I'll post back when I get more test results, but I'm at a loss here. Does DVDFab alter the profile based on certain parameters? Are there certain situations that could cause the encoder to end up with only half the target bitrate? I've never seen this before with other encoders.

Thanks,

Larry

[lalittle]

I just did the EXACT same tests I did yesterday, and I got totally different results. I used the same chapter of the same title of the same disc, and I used all the same encode settings, which are the same settings that I always use. Basically, it's just the default "mp4.h264.aac" in DVDFab, but I use 2-pass and change the target bitrate to 15,000Kbps. That's pretty much it -- I don't even know where I might find other variables to set.

This time, the encoder used the "high" profile (yesterday it used the "main" profile), and the average bitrate came out to 11,800Kbps -- less that I'd expect, but not the "half-target" bitrate I got yesterday.

I'll try doing the whole movie again and see what happens, but I just don't know what to think at this point.

Larry

[90312]

Gremlins

[lalittle]

Well -- I just got done encoding the entire film, and setting the target at 15,000Kbps resulted in a file having an overall average bitrate of 9,771Kbps. It did use the "high" profile instead of the "main" profile that the program decided to use yesterday, which I guess is a good thing, but it's rather distressing that the results are different even though I'm using the same settings.

It's not gremlins -- something in DVDFab 9's h.264 encoding process is definitely a bit screwy. On small samples it seems to get closer to the target bitrate, but with the full movie it's suspiciously off. Choosing "high" vs "main" profile seems to be somewhat random. I haven't seen the program encode with the "main" profile since my tests yesterday, but this DID happen yesterday.

I went back to my single chapter test, and I did get the same results I got earlier this evening, so at least TODAY it's being consistent. The encodes are still falling well short of the target bitrate, however, which shouldn't be happening.

Larry

PS. I've also had a few crashes just navigating the menus. I don't remember this happening recently with version 8.

[lalittle]

I have an important update to this.

I redid some tests, and things are at least consistant now -- the low bitrate output is repeatable, and I haven't seem the "Main profile" come back.

I tried another movie, however, and the mp4 output ended up very close to the target bitrate. This means that whatever is happening, it appears to be influenced by the movie itself. Perhaps something about the original encoding is tripping up the encoder, causing it to overestimate how many bits it will need, so it compresses more than it actually ought to. I would have thought that using 2-pass would prevent this from happening since it has a chance to "look at" the entire film before actually starting the encode, but the end result is still well below the target bitrate on certain films.

There is also still no explanation for why DVDFab chose the "Main" profile on a several occasions. I haven't seen it do this on my latest rips, but it did this several times the other day.

Is it possible that having DVDFab 8 on the system is causing some issues? I'm using the same output folder for my rips (and I believe for the temp files as well) -- is THIS possibly a problem? I wanted to keep things simple, and have all my rips in one spot, so I pointed to my existing DVDFab folder for these things. I did move the log output to a separate "DVDFab9\Logs" folder, but everything else points to the old location. I wouldn't think this could be a problem, but could the programmers confirm that this is okay?

Thanks,

Larry

[90312]

If quality is your thing why not try the passthrough? It will pass the video through to the mp4 container unchanged from the original in one pass.

[lalittle]

If quality is your thing why not try the passthrough? It will pass the video through to the mp4 container unchanged from the original in one pass.

Thanks again for the feedback here.

Unfortunately, using passthrough isn't an option for me with discs encoded with VC-1 because the next program in my pipeline can't read VC-1 files. I normally use a "copy" operation to create a de-copy-protected Blu-ray folder, which does not require any re-compression (similar to using "pass-through"), but I can only do this with discs using the AVC codec due, once again, to the current limitation of my pipeline.

Rather than change other parts of the pipeline, it's simply a lot easier to create re-compressed mp4 h.264 files from the VC-1 discs, which allows me to move forward using the same process that I do with the more prevalent AVC discs.

Larry

[lalittle]

I did another test with yet another VC-1 Blu-ray disc, and once again the bitrate ended up WELL below the target. My target was once again set at 15,000Kbps, but the resulting mp4 file's average video bitrate is only 6,756Kbps, less than half of the target. Some variance from the target bitrate is expected, but not nearly this much.

The mp4 was at least encoded with the "high" profile, but this bitrate issue is absolutely a problem, and a repeating pattern. It's somehow related to the movie itself (with some only some discs having the issue), as well as whether or not a single chapter or the entire movie is ripped/encoded.

Thanks for looking into this,

Larry

[lalittle]

Some theorizing:

I'm not clear how the DVDFab encoding algorithms work in specific situations, but is it possible that if an existing movie is encoded with a low enough quality, the encoder will determine that it simply doesn't "need" a bitrate as high as the target bitrate setting? In other words, are there circumstances where the encoder will simply decide to not use the entire target bitrate?

I haven't checked every disc yet, but one possible pattern here is that the mp4's that end up well below the target bitrate come from discs that have lower bitrates to start with. My theory is that in these circumstances, the encoder doesn't need the full target bitrate to achieve what it considers to be the highest possible quality, so the resulting file has a lower average bitrate than the target that was set.

What do you think?

Larry

[Wilson.Wang]

Hi,

I can confirm that it should be a bug.
Converted 1 chapter to mp4.h264.aac and set bitrate 15000, but the video came out just 9970, but near 14M in Fab8 (same source, same profile, same decoder and encoder).
Would you like to try converting to mkv?

Wilson