[MEncoder-users] Too many video packets in the buffer

[U. George]

On 06/18/2010 10:00 PM, Andrew Berg wrote:
> I doubt you'll find anything better than the default ffh264 in
> MEncoder).

At this moment, I am not finding anything that is working the way that I 
like it to.
For instance for a 30second *TEST* clip, with approx 898 frames, or some 
1797 fields I get this lastest analysis:

> ./mencoder -v -demuxer lavf  -noskip -oac copy -ovc x264 /home/gat/DIESEL/TEST_FP24/PF24_30Seconds.MTS -o foo.avi

> LAVF: build 3425280
> VIDEO:  [H264]  1920x1080  0bpp  59.940 fps    0.0 kbps ( 0.0 kbyte/s)
> [V] filefmt:35  fourcc:0x34363248  size:1920x1080  fps:59.940  ftime:=0.0167
=====================LATER===================
> Video stream: 22986.645 kbit/s  (2873330 B/s)  size: 43622427 bytes  15.182 secs  1830 frames
>
> Audio stream:  256.000 kbit/s  (32000 B/s)  size: 481280 bytes  15.040 secs

ftime is half of what it should be to reach 30 seconds. Frame count are 
double. This is all a bit confusing. THis is just not rock solid enough 
to try AVCHD video's of arbitrary camcorder settings.

Then there is "No Picture":
> Pos:   0.0s      2f ( 1%)   2fps Trem:   1min   0mb  A-V:-0.002 [0:0] A/Vms 3/379 D/B/S 0/1/0 ^M
> Pos:   0.0s      3f ( 1%)   2fps Trem:   1min   0mb  A-V:0.000 [0:0] A/Vms 3/330 D/B/S 0/2/0 ^M
> [h264 @ 0x8913da0]no picture
> Pos:   0.0s      4f ( 1%)   3fps Trem:   1min   0mb  A-V:0.002 [0:0] A/Vms 3/289 D/B/S 0/3/0 ^M
> Pos:   0.0s      5f ( 1%)   3fps Trem:   1min   0mb  A-V:0.003 [0:0] A/Vms 3/263 D/B/S 0/4/0 ^M
> [h264 @ 0x8913da0]no picture
> Pos:   0.0s      6f ( 1%)   3fps Trem:   2min   0mb  A-V:0.005 [0:0] A/Vms 3/239 D/B/S 0/5/0 ^M
> Pos:   0.0s      7f ( 1%)   4fps Trem:   2min   0mb  A-V:0.003 [0:0] A/Vms 3/229 D/B/S 0/6/0 ^M
> [h264 @ 0x8913da0]no picture

Then there is an unknown header ( where can I find AVCHD docs? ) :
> Unknown extra header dump: [0] [0] [0] [1] [9] [10] [0] [0] [0] [1] [27] [64] [0] [28] [ad] [0] [ec] [7] [80] [44] [7d] [e0] [22] [0] [0] [7] [d2] [0] [1] [d4] [c1] [c0] [40] [0] [a] [d2] [fe] [0] [0] [56] [97] [f3] [ef] [70] [60] [20] [0] [5] [69] [7f] [0] [0] [2b] [4b] [f9] [f7] [b8] [14] [0] [0] [0] [1] [28] [ea] [40] [13] [73] [90] [d2] [6b] [ac] [92] [11] [6] [93] [5d] [64] [a2] [42] [56] [ba] [ff] [d7] [ff] [eb] [f5] [ff] [af] [d5] [44] [12] [b5] [d7] [fe] [bf] [ff] [5f] [af] [fd] [7e] [bc]

AND I havent gotten into the inverse telecine filters yet!

If I try the -fps 30000/1001 i get different issues with "Skipping 
Frame" . I Dont want to add or subtact a frame what-so-ever!

What I'd like to try is to suggest to mencoder that the upper/lower 
fields should be combined into to one progressive field. Then field 
count would be correct,  and maybe timing ( 30 seconds ) would also be 
correct.

eventually I will try some of the other suggestions for getting it right 
( Right= avchd/PF24/30fps -> Avchd/24p/24fps of same video quality *AND* 
readable by Sony Vegas Pro )