HandBrake and Palm thread

Discussion in 'Multimedia (Palm OS)' started by Varjak, Jul 16, 2008.

Thread Status:
Not open for further replies.
  1. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    I upgraded to 0.9.3. Unfortunately, I won't really have time to play with it until I can solve my pTunes issues.
     
  2. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    Okay, I lied. I'm encoding my reference sample now. I used the basic iPhone Legacy setup and changed as little as possible. I moved the bitrate DOWN to 480. I chose 480 as my width but did not select a height (doesn't seem to be required, but MAY be specified if anamorphic is off). I did NOT select any anamorphic setting. I left the string in the advanced options as is for the iPhone AVC setting. Otherwise, I used the H.264 codec for an MP4 container file. I think I might have messed up the audio setup though. In the past, I used the AC3 stereo if possible, because it gives the best results for me (loud, crisp, clear). The AAC with HB mix resulted in a muddier, lower sound level. For some reason, AAC+AC3 no longer seems like an option (too bad, since it was the best option). The default for 0.9.3 suggested AAC with Dolby Pro Logic II. I don't think that's right.

    As I was doing this, I thought of something that may have helped your experiments. By using the relatively small dimensions of 480x208 (most of my encodings have been 480x272 or 480x256, which match the normal display aspect ratios of most programming); you cut out a lot of data per frame, which adds up quickly.
     
  3. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    One other thing. It's VERY helpful if you save your activity logs. I forget if this can be done automatically, but you can simply copy the info from the activity window and save it to a Word file or such. That way, we can each keep track of what works and what doesn't.
     
  4. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    As it processes, I can tell that AVC is much slower. It goes at about 1/2 the rate of the ffmpeg/MPEG4 codec. The first pass took 8 minutes for my 9 minute video and the second is projected at ~20 minutes, although it usually ends up lower than the estimate.
     
  5. raspabalsa

    raspabalsa Brain stuck BogoMipping

    Messages:
    7,264
    Likes Received:
    490
    Trophy Points:
    258
    Actually, no. When I tested T5 vs TX I used videos that had been encoded without the anamorphic option (selecting "None" in Anamorphic). I started playing with the anamorphic option later, after finishing the T5 vs TX tests.
    Again, no. After the initial surprise of having such lousy playback on the T5 I made sure our TCPMP and CorePlayer were set in exactly the same way. All plugins installed on both devices, and of the same versions too.

    Also remember that in his T5 CorePlayer performed better than TCPMP, while in my TX it was the opposite.
    Yes, I think that's also part of the problem. But as you say, it's surprising that similar devices will perform so differently.
    I don't know why anamorphic produces causes such performance hit, but it's so severe I don't think I'll use this option for the bulk of my videos. Even when lowering H.264 bitrate to 600 kbps I can't get sustained 24 fps frame rate. At most the TX can play at 20 fps, and at this frame rate I don't find palyback satistactory.
    No apparent difference at all. Both videos were properly formatted, but I didn't actually measure them to make sure. I may try this later, but I think they're the same because there is no apparent distortion.
    Yes, I can't make sense of HB's guide when it comes to anamorphic. I think I'll just keep anamorphic disabled, adjust video width to 480, and let HB handle the height. Doing so has produced very good results on both my TX and my computer.
    Not meaning to go into a discussion about FFMPEG vs XviD, but I have read the exact opposite: that XviD yields better quality than FFMPEG. This is why until now Iv'e encoded all my videos with XviD.
    I guess most of the LD's problems are because of the HDD and the fact that the "RAM" (the storage memory to be more accurate) is actually a partition of said drive. I can't talk fom experience since I've never used a LD.
    Actually TCPMP does not have buffering options, at least the last version (0.72 RC1) which I use. CorePlayer does have these options, but I'm not sure what the best settings are. I've noticed that opening a video on CP takes some time, presumably while CP is buffering data. Opening an H.264 video seems to take more time than an AVI of equal bitrate and size, but I haven't measured these times to be sure.
    I have played little with audio settings so far. I did encode a couple of MP4s with AC3 but couldn't get audio on CorePlayer. In fact, CP reported no audio stream in the files's properties. I didn't feel like pursuing an explanation, so I began encoding only with AAC and so far I like the results, although I guess at some point I may start testing various sound codecs and settings.
    This does help, for the reason you mention. But I don't think its that much of a difference. Remember that the extra pixels are only black bands, with little in the way of changes. H.264 with B-frames should take care of that quite nicely.
    Ah, that's a good idea, and one I had not thought about. I will start saving my encoding logs. Thanks!
    Yes, I've had similar time results. Very slow encoding, much slower than XviD. What's more, enabling the filters (interlacing, decombing, etc) slows down the process much more. For example, encoding the full Alexander movie (almost three hours of actual length) takes about 6 to 7 hours. But if I enable the Decomb filter the time more than doubles. Just the first pass took over 6 hours (with turbo first pass enabled), and then I decided to cancel the encoding because I needed to use the computer.

    I've read a lot of good things about the Decomb filter, and how it's best to leave it enabled always, but if the cost is such a slow encoding speed I won't use it often, if at all. This brings back a question I asked earlier: how to identify if a video needs decombing (or de-interlacing) a priori? Or would it be best (in terms of time) to simply encode it without these filters, then evaluate the result and re-encode with filters again if needed?
     
  6. raspabalsa

    raspabalsa Brain stuck BogoMipping

    Messages:
    7,264
    Likes Received:
    490
    Trophy Points:
    258
    I just finished my first full movie run test. I encoded the whole movie Alexander with H.264 @ 600 kbps and AAC @ 128 kbps. I did not use the anamorphic options, because at this point it's clear that my TX can't cope with the extra strain caused by these options, even when overclocked. Video frame size is 480 x 208. Frame rate is 24 fps. The file weighs 924MB.

    I played the whole movie (2h 48min) on a single uninterrupted run on CorePlayer with my TX overclocked to 520MHz. Playback was very good throughout the whole movie. Audio perfectly synced. Never did the TX fall behind. The video quality is excellent, much, much better than XviD at the same bitrate. Total played frames were 241512. Total dropped frames 144, which is 0.06% of the total.

    I think this test qualifies as "having a solution" :cool:

    I'm not sure I'll repeat this test at higher bitrates. 600 kbps looks so good even on my computer that I think higher bitrates are overkill. I do have a full encoded movie at 800 kbps, both with anamorphic on and off. I may run the test of the non-anamorphic version, but maybe a few days from now. I'm getting sick and tired of "Alexander"... I've watched it too many times this last week :D

    A test I may do is compare lower H.264 bitrates against high bitrate XviD samples, see at what bitrate H.264 yields a similar quality. However, I'm, not sure of the usefulness of this test. I'm perfectly comfortable with the high bitrate and large file size I'm using, and I'm not constrained by storage capacity or processing power (except when using anamorphic), so I don't need small file sizes and especially I don't want to sacrifice quality for storage.
     
  7. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    Raspy, I MEANT that the LACK of anamorphic made the difference between MY results and YOUR results, not between your TX and the T5.

    Also, what settings work for the TX might not be optimal for the T5. That's not really important anyway.

    You're wrong about the difference in pixels being the 'bands.' The whole point is that if my VIDEO is 480x272, that's the ACTIVE video window. It means the black bands are NARROWER! On any given frame, a 480x272 picture is 130,560 pixels and a 480x208 is 99,840. That's a 31% difference PER FRAME! The extra motion,etc. needs to be encoded; while the bars are most likely static and easily compressed, if they are even individually coded.

    You say in one part that you didn't measure the dimensions; but you always cite 480x208 as your aspect ratio. How do you know? Since you say you set 480 as the width and let HB set the height, I'm curious how to find out what the result is.

    In your most recent posts, you say the bitrate was 600. Did you encode these differently (setting a hard bitrate instead of shooting for a file size)? Do you use constant quality?

    It would be very helpful if you posted ALL your actual, specific settings as they appear on the HB control panel. That way I could do an actual comparison. The way you do it, changing more than one thing at a time, makes it difficult to see what each setting actually affects.

    I'm surprised you played Alexander on CorePlayer, since you said TCPMP actually outperforms it.
     
  8. raspabalsa

    raspabalsa Brain stuck BogoMipping

    Messages:
    7,264
    Likes Received:
    490
    Trophy Points:
    258
    Ah, well I mis-understood what you said. Anyway, I'll wait for the results of your encodings with 0.9.3.
    I don't think the optimal settings are that different for these two models. Same screen, same processor, similar hardware, same clock (in my tests with TX overclocked).
    Ok, I was thinking about my movie's aspect ratio, and somehow I thought that you were using the same ratio. I sometimes encode, say, a 480x208 movie to a 480x272 frame size, so the bands appear. I was thinking that you were using a video with same frame size and were doing a similar thing. Sorry fo the confusion.

    I didn't pay attention to the black bands until very recently, so I don't know how much they affect a movie's file size and performance.
    What I've been doing is defining width at 480 pixels, and leaving the height box empty, so HB can calculate it. This particular movie ends up having a height of 208 pixels (as reported by any player). I'm only citing this aspect (480 x 208) because so far I've only encoded one movie (Alexander).

    When I say "I didn't measure" I meant that I didn't measure the width and height of the image on the screen in milimeters to compare if there were differences between the anamorphic and the non-anamorphic videos.
    Yes, I now am settinga a defined bitrate. I've been playing with several bitrates, between 200 kbps and 900 kbps to compare quality. I'm no longer using fixed file size. In any case, it's easy to calculate file size based on video and audio bitrates, so I prefer to select th bitrates and estimate the file size.

    I'm not using constant quality. I believe that one of the strongest points of two-pass encoding is to measure bitrate requirements during the first pass, and set a variable bitrate for the final file accordingly. I think selecting constant quality would defeat that, unless I'm mistaken about how this option works.
    Yes, I should do that. Let me write down all the settings and report them later.
    Well, there are several reasons for doing this. CorePlayer has a more refined GUI. I don't mean the eye-candy, but the menu and tool layout. It's much easier to modify settings on CP. For example, if you want to modify colors in TCPMP you go to preferences -> video. But the settings windows occupies the whole screen. You have to adjust controls, go back to the movie, if not satsified go again to prefs, adjust, etc. On CorePlayer, you can call up the color controls right from the main window, and then the controls appear as a panel at the bottom of the screen, while the movie continues to play at the top. This is a most useful and nice feature. Several other tools have been improved as well.

    Another CP feature that I'm exploring is the buffer. I think that this may make CP perform as well as TCPMP if properly set. But I have no guidelines or manual. I've been playing with the buffer settings but I haven't found the ideal spot.

    Also, even though TPMP outperforms CP, playback on CP is perfectly adequate at the bitrates I'm using, so I may play on either player.

    There are many other adavantages of CP. Multiple audio streams, for example. This is not supported by TCPMP, which will only see the first audio stream in the file, and ignore any other.

    Last but not least, TCPMP is free, whereas CorePlayer cost me (and it wasn't cheap either). I like to see my money put to good use ;)
     
  9. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    Okay, although I converted this yesterday, I didn't put it on my LD until today. Obviously, the anamorphic setting is extremely demanding; because my 9 minute video played fine without anamorphic. I'll post the whole log below; but basically, as I think I stated above, I used the 'Legacy iPhone setting,; but moved the bitrate down to 480, no anamorphic, set the width to 480, left the height blank, no constant quality, left the advanced string as is in the 'advanced' setting, used AAC at the top quality with the Dolby 2 mix (I don't know how this got set, because when I called HB up later, it defaulted to something else), and no filters.

    The result wasn't really as good as my MPEG4 version. There were slight vertical bars to the left and right (so the picture wasn't quite as big). There seemed to be more pixelation, and the pixels seemed to be bigger. The resolution and color did not seem to be as good and there was DEFINITELY more artifacting. To be fair, I did it at a bitrate of 480, so there could be some improvement there. According to the media info, the dimensions were 480x288! and 29.97 fps, with no frames dropped. The log indicates there was some cropping, but I didn't configure that.

    I'd love to find out more about the anamorphic settings (what they actually do, etc.); but unfortunately I don't think we'll be able to get any help with this on the HB forum. That's what got me screwed up there in the first place.

    Here's the log:

    ### CLI Query: -i "D:\VIDEO_TS" -t 2 -c 1 -o "C:\Program Files\palmOne\HandBrake conversions\Rendezvous AVC2.mp4" -f mp4 -X 480 -e x264 -b 480 -2 -T -a 1 -E faac -B 128 -R 48 -6 dpl2 -D 1 --markers="C:\Documents and Settings\XXXX\Local Settings\Temp\VIDEO_TS-2-chapters.csv" -x level=30:cabac=0:ref=1:analyse=all:me=umh:no-fast-pskip=1 -v
    #########################################
    [18:40:10] hb_init: checking cpu count
    [18:40:10] hb_init: starting libhb thread
    HandBrake 0.9.3 (2008112300) - http://handbrake.fr/
    1 CPU detected
    Opening D:\VIDEO_TS...
    [18:40:10] hb_scan: path=D:\VIDEO_TS, title_index=2
    [18:40:10] scan: trying to open with libdvdread
    [18:40:10] scan: DVD has 12 title(s)
    [18:40:10] scan: scanning title 2
    [18:40:10] scan: opening IFO for VTS 2
    [18:40:10] pgc_id: 1, pgn: 1: pgc: 0x1584050
    [18:40:10] scan: vts=2, ttn=1, cells=0->0, blocks=0->270342, 270343 blocks
    [18:40:10] scan: duration is 00:08:39 (519133 ms)
    [18:40:10] scan: checking audio 1
    [18:40:10] scan: id=80bd, lang=Unknown (AC3), 3cc=und ext=0
    [18:40:10] scan: checking audio 2
    [18:40:10] scan: id=81bd, lang=Unknown (AC3), 3cc=und ext=0
    [18:40:10] scan: checking audio 3
    [18:40:10] scan: id=82bd, lang=Unknown (AC3), 3cc=und ext=0
    [18:40:10] scan: title 2 has 1 chapters
    [18:40:10] scan: chap 1 c=0->0, b=0->270342 (270343), 519132 ms
    [18:40:10] scan: aspect = 0
    [18:40:10] scan: decoding previews for title 2
    [18:40:10] scan: audio 0x80bd: AC-3, rate=48000Hz, bitrate=224000 Unknown (AC3) (2.0 ch)
    [18:40:10] scan: audio 0x81bd: AC-3, rate=48000Hz, bitrate=448000 Unknown (AC3) (5.1 ch)
    [18:40:10] scan: audio 0x82bd: AC-3, rate=48000Hz, bitrate=224000 Unknown (AC3) (Dolby Surround)
    Scanning title 2...
    Scanning title 2...
    [18:40:10] scan: 10 previews, 720x480, 29.970 fps, autocrop = 50/50/0/0, aspect 4:3, PAR 8:9
    [18:40:10] scan: title (0) job->width:640, job->height:384
    [18:40:11] libhb: scan thread found 1 valid title(s)
    + title 2:
    + vts 2, ttn 1, cells 0->0 (270343 blocks)
    + duration: 00:08:39
    + size: 720x480, aspect: 1.33, 29.970 fps
    + autocrop: 50/50/0/0
    + chapters:
    + 1: cells 0->0, 270343 blocks, duration 00:08:39
    + audio tracks:
    + 1, Unknown (AC3) (2.0 ch), 48000Hz, 224000bps
    + 2, Unknown (AC3) (5.1 ch), 48000Hz, 448000bps
    + 3, Unknown (AC3) (Dolby Surround), 48000Hz, 224000bps
    + subtitle tracks:
    Reading chapter markers from file C:\Documents and Settings\XXXX\Local Settings\Temp\VIDEO_TS-2-chapters.csv
    Modified x264 options for pass 1 to append turbo options: level=30:cabac=0:ref=1:analyse=all:me=umh:no-fast-pskip=1:ref=1:subme=1:me=dia:analyse=none:trellis=0:no-fast-pskip=0:8x8dct=0:weightb=0
    [18:40:11] 2 job(s) to process
    [18:40:11] starting job
    [18:40:11] Width out of bounds, scaling down to 480
    [18:40:11] New dimensions 480 * 288
    [18:40:11] work: sanitizing track 0 mixdown Dolby Pro Logic II to Stereo
    [18:40:11] job configuration:
    [18:40:11] * source
    [18:40:11] + D:\VIDEO_TS
    [18:40:11] + title 2, chapter(s) 1 to 1
    [18:40:11] * destination
    [18:40:11] + C:\Program Files\palmOne\HandBrake conversions\Rendezvous AVC2.mp4
    [18:40:11] + container: MPEG-4 (.mp4 and .m4v)
    [18:40:11] + chapter markers
    [18:40:11] * video track
    [18:40:11] + decoder: mpeg2
    [18:40:11] + bitrate 7500 kbps
    [18:40:11] + frame rate: same as source (around 29.970 fps)
    [18:40:11] + dimensions: 720 * 480 -> 480 * 288, crop 50/50/0/0
    [18:40:11] + encoder: x264
    [18:40:11] + options: level=30:cabac=0:ref=1:analyse=all:me=umh:no-fast-pskip=1:ref=1:subme=1:me=dia:analyse=none:trellis=0:no-fast-pskip=0:8x8dct=0:weightb=0
    [18:40:11] + bitrate: 480 kbps, pass: 1
    [18:40:11] * audio track 0
    [18:40:11] + decoder: Unknown (AC3) (2.0 ch) (track 1, id 80bd)
    [18:40:11] + bitrate: 224 kbps, samplerate: 48000 Hz
    [18:40:11] + mixdown: Stereo
    [18:40:11] + encoder: faac
    [18:40:11] + bitrate: 128 kbps, samplerate: 48000 Hz
    [18:40:11] reader: first SCR 1265
    [18:40:11] encx264: keyint-min: 30, keyint-max: 300
    x264 [info]: using cpu capabilities: MMX2 SSE2 Cache64
    x264 [info]: profile Baseline, level 3.0
    No accelerated IMDCT transform found
    [18:40:11] sync: expecting 15588 video frames
    [18:40:14] mpeg2: "Chapter 1" (1) at frame 0 time 3003
    [18:40:14] sync: first pts is 3003
    [18:53:26] reader: done. 0 scr changes
    [18:53:26] sync: got 15571 frames, 15588 expected
    [18:53:26] work: average encoding speed for job is 19.642456 fps
    [18:53:26] mpeg2 done: 15572 frames
    [18:53:26] render: lost time: 0 (0 frames)
    [18:53:26] render: gained time: 0 (0 frames) (0 not accounted for)
    x264 [info]: slice I:52 Avg QP:21.57 size: 7125 PSNR Mean Y:43.16 U:49.05 V:48.54 Avg:44.14 Global:43.55
    x264 [info]: slice P:15519 Avg QP:24.56 size: 1966 PSNR Mean Y:40.41 U:47.99 V:47.40 Avg:41.60 Global:40.94
    x264 [info]: mb I I16..4: 52.5% 0.0% 47.5%
    x264 [info]: mb P I16..4: 23.2% 0.0% 0.0% P16..4: 45.3% 0.0% 0.0% 0.0% 0.0% skip:31.6%
    x264 [info]: final ratefactor: 27.36
    x264 [info]: SSIM Mean Y:0.9608208
    x264 [info]: PSNR Mean Y:40.414 U:47.997 V:47.399 Avg:41.610 Global:40.947 kb/s:475.51
    [18:53:27] starting job
    [18:53:27] Width out of bounds, scaling down to 480
    [18:53:27] New dimensions 480 * 288
    [18:53:27] work: sanitizing track 0 mixdown Dolby Pro Logic II to Stereo
    [18:53:27] job configuration:
    [18:53:27] * source
    [18:53:27] + D:\VIDEO_TS
    [18:53:27] + title 2, chapter(s) 1 to 1
    [18:53:27] * destination
    [18:53:27] + C:\Program Files\palmOne\HandBrake conversions\Rendezvous AVC2.mp4
    [18:53:27] + container: MPEG-4 (.mp4 and .m4v)
    [18:53:27] + chapter markers
    [18:53:27] * video track
    [18:53:27] + decoder: mpeg2
    [18:53:27] + bitrate 7500 kbps
    [18:53:27] + frame rate: same as source (around 29.970 fps)
    [18:53:27] + dimensions: 720 * 480 -> 480 * 288, crop 50/50/0/0
    [18:53:27] + encoder: x264
    [18:53:27] + options: level=30:cabac=0:ref=1:analyse=all:me=umh:no-fast-pskip=1
    [18:53:27] + bitrate: 480 kbps, pass: 2
    [18:53:27] * audio track 0
    [18:53:27] + decoder: Unknown (AC3) (2.0 ch) (track 1, id 80bd)
    [18:53:27] + bitrate: 224 kbps, samplerate: 48000 Hz
    [18:53:27] + mixdown: Stereo
    [18:53:27] + encoder: faac
    [18:53:27] + bitrate: 128 kbps, samplerate: 48000 Hz
    [18:53:27] encx264: keyint-min: 30, keyint-max: 300
    x264 [info]: using cpu capabilities: MMX2 SSE2 Cache64
    [18:53:27] reader: first SCR 1265
    [18:53:27] mpeg2: "Chapter 1" (1) at frame 0 time 3003
    x264 [info]: profile Baseline, level 3.0
    No accelerated IMDCT transform found
    [18:53:31] sync: expecting 15588 video frames
    [18:53:31] sync: first pts is 3003
    [19:18:50] sync: got 15571 frames, 15588 expected
    [19:18:50] work: average encoding speed for job is 0.000000 fps
    [19:18:50] reader: done. 0 scr changes
    [19:18:52] mpeg2 done: 15572 frames
    [19:18:52] render: lost time: 0 (0 frames)
    [19:18:52] render: gained time: 0 (0 frames) (0 not accounted for)
    x264 [info]: slice I:52 Avg QP:21.72 size: 7201 PSNR Mean Y:43.12 U:49.58 V:49.12 Avg:44.19 Global:43.65
    x264 [info]: slice P:15519 Avg QP:24.51 size: 1983 PSNR Mean Y:40.73 U:48.48 V:47.94 Avg:41.95 Global:41.47
    x264 [info]: mb I I16..4: 33.1% 0.0% 66.9%
    x264 [info]: mb P I16..4: 3.7% 0.0% 2.1% P16..4: 59.8% 15.7% 3.5% 0.1% 0.1% skip:15.1%
    x264 [info]: SSIM Mean Y:0.9628950
    x264 [info]: PSNR Mean Y:40.740 U:48.481 V:47.945 Avg:41.955 Global:41.476 kb/s:479.52
    [19:18:52] libhb: work result = 0
    Rip done!

    ############ End of Log ##############
     
  10. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    One other thing, perhaps v. 0.9.3 IS superior.
     
  11. FireNWater

    FireNWater Mobile Deity

    Messages:
    108
    Likes Received:
    0
    Trophy Points:
    21
    Hi All,

    I haven't figure out how to make the nice quote boxes so please forgive me for just texting. I also compose offline on notepad and then just copy over post. I've listed the Post # for reference.

    Post #30 excerpts from Razzy
    "In all four clips I selected "None" in the "anamorphic" box. "
    Razz's tests were of non-anamorphic from the first revision of this thread.

    "The main improvement is that the square blocks common to the XviD codec are nowhere to be seen. "

    Post #56? 08-03-2009 05:37 PM
    "how to identify if a video needs decombing (or de-interlacing) a priori?

    Or would it be best (in terms of time) to simply encode it without these filters, then evaluate the result and re-encode with filters again if needed?"
    ***********
    The artifacts supposedly indicate interlacing issues. However, if your new H.264 files do not have it, then I don't think you need to de-interlace/comb anything.
    Ask yourself the source of the video? If recording analog TV signal to begin with, then it is likely interlaced. If it was broadcasted, then likely interlaced.
    If your vid looks good then I wouldn't bother turning it on since the source is probably not interlaced to begin with and if they were, it seems that the higher quality H.264 is dealing with it w/o requiring decombing.

    Post #65: XviD yields better quality than FFMPEG
    yes, d'accord. HB's attitude is to value speed over quality hence their eventually dumping Xvid support.
    But like you said, we can still continue to use the older stuff.

    Anamorphic or not?

    If the source video says it's Anamorphic, then HB's anamorphic features could be used to advantage. This is listed on DVD specs now if it is. If it says Widescreen or Standard Widescreen, then the DVD is NOT anamorphic and its no use using the anamorphic HB features.

    Testing files: when I used to interview for production and had to submit videos, the interviewers said no more than 3 min and make it relevant, that is, skip the black lead ins which were necessary when broadcasting on web. If people are going to submit vids for review/testing, it might be a good idea to also keep to the 3Min limit even though your bandwidths Might accommodate more. Make it easy on both ends for review/test. That 3 min should count for whatever you are testing - motion, audio, colors, etc. There are free video editors to chop the sections you need if your file is too large.

    For HB, once it starts encoding, I found I could simply preview by waiting till a few minutes were encoded and saving that file under a different name.

    I could then take a look at with the VLC player or other to test if the sound/video was coming thru the way I wanted or if I even got the correct file to begin with.
    Hard to tell when everything is numbered hence the quick checks before letting the HB do ALL of the file. Even if a player told me which chapter and file, it didn't always correspond to the DVD files. IF it was the correct file, I'd let HB continue. If not, then I'd abort and go test another till I got something that played well on TCPMP.

    I think this is one way you can check for artifacts before encoding 30 plus minutes. If they are present, they would show for the first 3 min as much as the last 45 or 120 min of the movie.

    Maybe you already do this. Just listing that tip here in case for future newbies that this is how we check quickly in the beginning with HB files.

    That's how I figured out the various codecs/audio/video specs that would work for me.
    I did not know each codec in detail but guessed based on Pocket ENcoder and other video specs that were already completed and used them for reference. The rest I've learned from observing these threads and now thru HB's super detailed Wicki.

    My goal was simply to get a file that would run on my TX with the sound and video. Thanks to your specs I have some approach to getting something in H.264 to run eventually.:+)
     
  12. raspabalsa

    raspabalsa Brain stuck BogoMipping

    Messages:
    7,264
    Likes Received:
    490
    Trophy Points:
    258
    480 kbps is lower than I think ideal for H.264. I encoded some samples down to 200 kbps, but I think I'll settle for 600 kbps, which results in much better quality than my 600 kbps XviD videos.

    Varjak, do know what the differences are between the profiles? I've using only the iPhone & iPod Touch profile. I think your test resulted in lower quality than MPEG4 partly because of the profile you chose. It must be labeled as "Legacy" for some valid reason, I think. Perhaps it disables some encoder features to allow it to be played on older devices?

    What bitrate did you use on the FFMPEG video you used as comparison? The same one (480)?
    Do you have any idea why this happened? Looks like 50 pixels were cropped both left and right, right?
    The bars you mention are someting I haven't seen in my encodings. As for pixelation and artifacting, I see much less on my H.264 than on my XviDs.

    Here's the detail of the settings I've used for what I consider are the best results I've got so far:

    Code:
    Profile: iPhone & iPod Touch
    Format: MP4
    Crop: Automatic
    Width: 480
    Height: Left blank
    Anamorphic: None
    Filters: None enabled
    Video: H.264
    Framerate: Same as source
    Grayscale encoding: Unchecked
    Two-pass encoding: Checked
    Turbo First pass: Checked
    Target size: Left blank
    Average bitrate: 600 to 800
    Constant Quality: I've never moved this slider. However, when I select 600 kbps, I later find that it's automatically moved to 59
    Audio: AAC
    Mixdown: Dolby Pro-Logic II (default, I haven't changed it)
    Samplerate: 48 kHz
    Bitrate: 128
    DRC: 1 (default, I haven't changed it)
    
    Under advanced H.264 Options:
    Reference frames: 2 (default)
    Mixed references: Checked (default)
    B-Frames: 2
    Direct prediction: Default (spatial)
    Weighted B-Frames: Checked
    Pyramidal B-Frames: Unchecked
    The other options at the right of the advanced tab are at their defaults. I don't know yet how to use them.
    Yeah, I'd like to know more about anamorphic, how it works and why it affects playback on our devices. Since the HB forum is not a friendly place I'm thinking about visiting CorePlayer's forum. They may have some answers, especially about the decoding and why it's hit so hard when enabling anamorphic.
     
  13. FireNWater

    FireNWater Mobile Deity

    Messages:
    108
    Likes Received:
    0
    Trophy Points:
    21
    and slightly off topic,
    Many DVDs now bundle with a Ipod playable version which saves having to convert for a PMP but they are in H.264, not Xvid/Avis that can run on our Tx/TCPMPs.

    I just saw on ARS that future DVD bundles will provide the movie on micro SD cards for about $11 US additional in future.

    Still, all of them are in modern formats so I think I still have to rip to TX/TCPMP friendly files.
    I got a DVD with a Ipod friendly file but still couldn't play it on TX.
    So I need to learn how to convert it myself but there are a lot of people out there who don't know how to convert and would go for the convenience.
     
  14. raspabalsa

    raspabalsa Brain stuck BogoMipping

    Messages:
    7,264
    Likes Received:
    490
    Trophy Points:
    258
    Hi FNW! To create quotes, it's easiest to select the whole post(s) to be quoted. If you want to quote a single post, just click on the "Quote" button at the bottom of the post. This will open the advanced reply window. There you can edit the quote and there you will also see what code and format are used to create a quote, so you can start getting familiar with it.

    If you want to quote multiple posts (like you just did), then click on the small button with the " sign, right next to the "Quote" button. This is the multi-quote button. You can select several posts (by clicking on their respective multi-quote buttons), and you can even select posts from multiple pages (at least pn Firefox I can do that). Once you've selected all of the posts, click on the "Post Reply" button at the bottom left of the page. A new reply window will open where you will see all the quotes. There you can edit and compose your reply.
    Ah, that's interesting info. Perhaps I should try some encodings on PocketDivX Encoder, Fairuse, or Handbrake with the de-interlace options enabled, see if the output improves.
    Hmmm.... but what about DVDs? Is it sagfe to assume that DVDs of newer movies are more likely to contain non-interlaced videos (because they're likely to be played on LCDs, Plasmas, etc), while DVDs of older movies may be interlaced because back then only CRTs were used for playback?
    Yes, except that HB's manual recommends to keep "decomb" enabled for better quality. The problem is that encoding with decomb takes much longer than without it.
    But HB isn't dropping XviD. It's dropping AVI (the container). You can still use XviD in an MP4 container.
    The problem is that if we use anamorphic our devices can't keep up during playback. And I still don't understand what's the advantage of using anamorphic.
    That makes sense. You worked in video production?
    I've done the same when encoding with PDE. The problem is that you're seeing the output of the first pass, meaning output quality is lower than on the final video (second pass). It can give you an idea of how good the final will be, but it's not an accurate result. But it does help to verify you selected correct video or correct stream (audio, subtitle, etc).
     
  15. raspabalsa

    raspabalsa Brain stuck BogoMipping

    Messages:
    7,264
    Likes Received:
    490
    Trophy Points:
    258
    But TCPMP can play H.264, that's the whole point of this exercise. I don't understand what you mean above.
    Interesting, but too expensive IMO, and may not be encoded at the settings one prefers.
    FNW, probably you haven't installed the AVC codec for TCPMP. Check with Filez or another file manager. If this codec is not installed on your TX, download and install it. I don't have a direct link, but search the forum's member list for jigwashere and see his profile. In his signature he has links for both the AVC and AAC codec for TCPMP.
     
  16. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    Raspy:

    1. Where is the- " sign, right next to the "Quote" button ? I don't have that next to my quote button, I do have 'multiquote' but I don't think that was there until recently. I know how to quote whole posts (and I sometimes manually delete irrelevant 'graphs); but I always wondered who to multi-quote.

    2. I don't know what the diff. is b/t the legacy iPhone and the iPhone. I just figured that the older iPhone settings would be closer to the Palm then the newer settings meant for the faster, upgraded iPhone.

    2. I deliberately chose a low bitrate to see if I could get a decent running file. As I recall, the original HB settings recommended bitrates as low as 250 for converting AVC files.

    3. You keep trying to compare MPEG4 bitrates to AVC bitrates. That's like comparing the clock speeds of a 286 computer to the latest Pentium. Not relevant. They are different standards.

    4. I'm reluctant to use b-frames; because I think it's been posted before that the basic form of AVC on the Palm doesn't support it. Until someone actually can explain to me what it does, I'll avoid it. I looked it up on Wikipedia, and it seems to refer to compression, which means it might slow down how the Palm processes video. It probably DOES make the file smaller though.

    5. I'm interested in the interlacing too. I would think that would only matter if the content CAME from an interlaced source (i.e. taken off the TV for example). I don't think sources like DVD's are interlaced in and of themselves. Interlacing relates to the DECODING and PROJECTION technology (such as the diff. b/t 'regular' DVD players and progressive scan DVD players). The same DVD would be read and displayed differently by each of these DVD players.

    6. Raspy, in your advanced settings, did your iPhone default put any kind of string of terms in the window at the bottom of the control panel?
     
  17. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    FNW, can you write more about anamorphic? In my experience, most recent DVD's (even of older movies) are done in anamorphic video (I tend to avoid pan-and-scan versions of movies like the Plague). I think you're right about Standard Widescreen (which I haven't seen much of recently); but I think most things labeled 'widescreen' ARE anamorphic.

    My understanding of anamorphic (with regard to optics) is that a special lens is used when FILMING that compresses the image down so it fits on standard 35 mm film. To project the finished film, another lens re-expands the image for projections. How this relates to video and video encoding is beyond me. I thought it was about best achieving the original aspect ratio and dimensions of the original source film. The HB seems to express it is that all DVD's have standard frame sizes with a 'flag' that tells the device what aspect ratio to use when displaying that image. It's like a multiplier. So I don't know how, without anamorphic, the encoder will know the proper dimensions for the final product. With my reference sample, the non-anamorphic file came to 480x288. The anamorphic MP4 file encoded with MPEG4 (ffmpeg) was 480x256. Anything you could add would be VERY much appreciated.

    I disagree with your supposition that artifacts will just as likely appear in the beginning, as later. Not in my samples. I use a 9 minute reference sample and the artifacting appeared only in the most demanding parts. If you don't have an action scene or something else that might be demanding to the encoder until midway through the source material, you may not see artifacting.

    You mention blocky areas with XviD. That's what I got with AVC. I don't know if a higher bitrate will solve that; but that's next on the list of experiments.
     
  18. raspabalsa

    raspabalsa Brain stuck BogoMipping

    Messages:
    7,264
    Likes Received:
    490
    Trophy Points:
    258
    There are three buttons at the bottom right corner of every post. From left to right: Quote, Multi-Quote, and Reply. The multi-quote has a " and a + in its icon. Just clic the multi-quote button on everty post you want to quote. When done, clic the Post Reply bottom at the bottom left of the page.
    Perhaps the Palms are not as obsolete as we think, as my TX is able to play videos encoded with the newer iPhone/iPod profile. Is there some place where we can find what are the differences between profiles?
    That bitrate may be good for slow movies (dramas, comedies) or movies not at widescreen, but I doubt you'll get good results for a widescreen action movie at such low bitrate.
    Not relevant? It was you who observed that my bitrate comparisons were unfair and suggested a differet (but still bitrate vs bitrate) comparison. Let me quote:
    What's more, in my opinion a bitrate comparison between H.264 and XviD (or FFMPEG) is very valid. It's true that both codecs handle video data in different ways, but the end result is a stream of data in a container. My comparisons aim to find an ideal balance between video quality and file size. By comparing against my XviD results I get an idea of how much I can shrink the final file without compromising quality, with the XviD quality (at the high bitrate settings I've used) serving as a floor value. In the end I don't care how or what the codec does to the data, I'm just interestend in the final result.
    Would be nice to have a confirmation about this. But even if Palm's AVC codec doesn't support B-Frames, I'll still use them, because my videos are not only used on the Palm, but on other devices as well, some of which do support B-Frames.

    My interpretation of B-Frames is that they are "partial" frames (meaning they don't contain all the pixels of a frame) that are interpolated between two actual frames, based on the encoder's analysis of the video data. In essence, the encoder predicts what the next frame will look like and create a B-Frame based on that. But I'm not yet familiar with all they do or how.
    Yes, for the settings I mentioned above this is what appears there:

    level=30:cabac=0:ref=2:mixed-refs=1:analyse=all:me=umh:no-fast-pskip=1:bframes=2:weightb=1

    Can you make sense of all of this?
    Not exactly. The image is not compressed, but stretched. 35mm film has an aspect ratio of about 4/3. This means that a widescreen movie will not use the whole height of the film and will have lower resolution than one that fully uses the whole height. The anamorphic lens stretches the image vertically so it will use the whole height of the film and thus the image will be full resolution. The projector will use other lenses that squash the image vertically to show it at the correct aspect ratio.
    There we agree completely. Further, my anamorphic and non-anamorphic videos look exactly the same on both my TX (disregarding the fact that the TX can't cope with the processor demand of the anamorphic version) and computer, so I don't get what the advantage is.
    I definitely get no blocky areas with H.264 even at bitrates as low as 200 kbps for the test video I've been using. But I do get some blocky areas with XviD, even at 450 kbps for the same video. They're not too marked and appear just a few times, though.

    Some of my older encodings with XviD do have lots of blocky areas, especially at the low bitrates (300, 350) I used when I started encoding, years ago.
     
  19. Varjak

    Varjak Mobile Deity

    Messages:
    6,428
    Likes Received:
    35
    Trophy Points:
    0
    1. I don't have those marks in my 'multiquote' button; and my multiquote button doesn't even function. It just turns greenish; but nothing happens.

    2. My point about bitrates is not that different trials aren't useful; but that you seem to equate a particular bitrate on AVC with a particular bitrate on MPEG4. I don't know what the 'equivalency' should be; but it's not to compare 600 AVC with 600 MPEG4. That's all. There's probably an optimum bitrate for each codec, and they are probably wildly different (maybe 2:1 or something). The two different codecs do give you tradeoffs in file size v. quality. You haven't done any (or maybe many) MPEG4 encodes, while I'm going to avoid XviD. You and I seem to have different experiences. For me, MPEG4 give noticeably better quality, and with ever-large SDHC cards, space is less of an issue than quality and playability. Not to mention encoding times, which are much longer for AVC.

    3. I don't know why all of a sudden you seem to be an expert on anamorphic; but anamorphic does not stretch the image (horizontally or vertically). First, with FILM, it squeezes the image down, and then the companion lens re-expands it for projection. You may be talking exclusively about anamorphic with respect to DVD anamorphic widescreen; but even then, it is using a similar technique to squeeze a widescreen image to a 4:3 frame and then when it is played, to re-expand it to its natural width. No stretching.

    FILM:

    http://en.wikipedia.org/wiki/Anamorphic#Lens_makers_and_corporate_trademarks

    DVD:

    http://en.wikipedia.org/wiki/Anamorphic_widescreen

    I'm not looking for an argument, but I don't want to introduce flawed information into what's already a complicated discussion.

    4. Good point about b-frames. That was my initial thought as well, since I expected to get an iPhone. But if I remember correctly, the iPhone USED anamorphic, which is why I never really tried it without it. Does the 'new' iPhone setting NOT use anamorphic? I may have gotten the impression it did from the 'help' forums.

    5. I'm glad you could grab the 'advanced' line from HB. I tried to copy and paste it and it wouldn't work.
     
  20. FireNWater

    FireNWater Mobile Deity

    Messages:
    108
    Likes Received:
    0
    Trophy Points:
    21
    Hi all, been busy with classes and out of the loop.
    turns out, there is always room to learn new stuff.

    wow , what a difference a year makes.
    I went back to the beginning of this thread and saw my own problems in having TCPMP run a file with a mp4 container using a H.264/avc Video codec and AAC audio codec last year.

    This year, I tried the config Raz listed.
    Added the AAC codec from Jig's sig. Did not add the AVC yet since I have to register for Treo site first.

    First couple of files, not smooth, not good. But Raz's success told me it's doable. Not so sure I want to overclock just yet.

    So back to a source DVD for video. Coded it w/AVC/AAC, stuck it in the m4v
    mp4 is HB's default.
    chose a Quicktime container m4v, v for video I think if your file has chapter markings.
    Found HB continues to code m4v (letter v) no matter if I pick mp4 default (letter p). It's another setting if you want to absolutely have the mp4 extension tagged on the file.

    * added the AAC codec. After more trial and error, it actually ran smooth!!!!! So yippeeee! Thanks guys!

    My experiment: only coded the shortest chapter possible from a movie DVD. I've coded half the movie and 30 min runs no problem and NO overclocking required for this file in mp4.

    digression: So yeahhhh, I'm thinking Wowza and of Woz the Oz the too cuz , well, just cuz. ok.

    Next experiment was some mp4 files from a youtube source.
    First downloads were choppy, only seeing keyframes, no motion, audio fine though. Looking at source, it was 780 px (desksize) so back to HB to resize it down to 480.
    At least the TX could try to play it and then got a new message - out of memory. Hence learning look at the file dimensions before putting it in.

    Learned to code a pre-set so I don't have to set it each time.
    After a few more experiments, got a 50,000 kb file down to 21,000 file and this 480 px wide file runs smooth on TX.

    It doesn't always work to try to size it down but when i can, I get to play it in the TX, no additional boosting or weird contortions required.

    And with Download helper updated on FF, high quality mp4s are now avail for download. no conversion required. If poster put them at less than 480 px to begin with, then the download can run immediately on the TX and like Razzy pointed at the revival, omg, the mp4s are really a step above in quality.

    I'm definitely saving in mp4 whenever possible from now on.

    At first revival, I thought it was impossible to play mp4s without overclocking. Now I realize from further research, that it's for files with a lot of action (Razzy's Alex) or for 780 size (desk size files). Some other poster mentioned he doesn't want to have to keep 2 versions (TX friendly versus desk size) and thus overclocks to view.
    Razzy's were sized for TX but I guess because of the action, it takes more oomph to run. Now I realize with low demand movies, TX size files will run just fine without overclocking.

    So thank you guys for a spirited discussion and motivating me to learn new stuff.
     
Thread Status:
Not open for further replies.

Share This Page