Re: [FFmpeg-user] flags +ildct+ilme produce corrupted file
Massimo Battistel <battistel <at> gmail.com> writes:
1. How to know the exact differencies between the 2 presets?
Did you already try to reproduce your issue with the x264 binary? If it is reproducible with x264, you could ask on the x264 mailing list, if it is not reproducible with x264, there is probably a bug in FFmpeg which we will try to fix.
2. Is there any relation between my issue and #993? ( http://trac.ffmpeg.org/ticket/993)
Only if you problem is not reproducible with -threads 1 Carl Eugen
1. How to know the exact differencies between the 2 presets?
Did you already try to reproduce your issue with the x264 binary? If it is reproducible with x264, you could ask on the x264 mailing list, if it is not reproducible with x264, there is probably a bug in FFmpeg which we will try to fix.
I tried this command line: x264 --preset ultrafast -o out.h264 ref.mp4 Not sure it is exactly like the one with ffmpeg, but output video is fine, no errors. I think I can't add field order infos, since it generates just a video track.
2. Is there any relation between my issue and #993? ( http://trac.ffmpeg.org/ticket/993)
Only if you problem is not reproducible with -threads 1
The problem persists with -threads 1 (and ffmpeg command line). -- Massimo Battistel.
Massimo Battistel <battistel <at> gmail.com> writes:
1. How to know the exact differencies between the 2 presets?
Did you already try to reproduce your issue with the x264 binary? If it is reproducible with x264, you could ask on the x264 mailing list, if it is not reproducible with x264, there is probably a bug in FFmpeg which we will try to fix.
I tried this command line:
x264 --preset ultrafast -o out.h264 ref.mp4
Is the original problem with ffmpeg also reproducible without flags +ildct+ilme?
Not sure it is exactly like the one with ffmpeg,
but output video is fine, no errors. I think I can't add field order infos
Is the original problem with ffmpeg not reproducible with -top 0 (or without top)? Carl Eugen
Is the original problem with ffmpeg also reproducible without flags +ildct+ilme?
No, without +ildct+ilme encoding is fine with no errors.
Is the original problem with ffmpeg not reproducible with -top 0 (or without top)?
-top 1, -top 0 or without top gives always the same encoding issue with the same errors ("top block unavailable for requested intra mode at 0 0") thanks, MB
2014-03-05 16:23 GMT+01:00 Massimo Battistel <battistel@gmail.com>:
Is the original problem with ffmpeg also reproducible without flags +ildct+ilme?
No, without +ildct+ilme encoding is fine with no errors.
Is the original problem with ffmpeg not reproducible with -top 0 (or without top)?
-top 1, -top 0 or without top gives always the same encoding issue with the same errors ("top block unavailable for requested intra mode at 0 0")
thanks, MB
Any news? Do you think this is a libx264 issue or ffmpeg issue? Do you need further tests? I have another interesting information: the issue happens only when recorded video starts with a some solid colors. With these yuv420 colors issue happens: y=0,u=v=128 (black) y=16,u=v=128 (very dark gray) y=0,u=v=0 (green) With this color there is no issue, no errors: y=128,u=v=128 (grey) thanks, MB
Massimo Battistel <battistel <at> gmail.com> writes:
Do you think this is a libx264 issue or ffmpeg issue? Do you need further tests?
Did you already test interlaced encoding with x264? Carl Eugen
Did you already test interlaced encoding with x264?
I tried the following: x264 --preset ultrafast --interlaced --tff -o out.h264 ref.mp4 No errors, no artifacts. But I'm not really expert about x264 command line. If someone could suggest the exact equivalent of "-flags +ildct+ilme", I could make a better test. thanks, MB
2014-03-06 17:17 GMT+01:00 Massimo Battistel <battistel@gmail.com>:
Did you already test interlaced encoding with x264?
I tried the following:
x264 --preset ultrafast --interlaced --tff -o out.h264 ref.mp4
No errors, no artifacts. But I'm not really expert about x264 command line. If someone could suggest the exact equivalent of "-flags +ildct+ilme", I could make a better test.
Is the following the correct x264 mail list? https://mailman.videolan.org/listinfo/x264-devel thanks, MB
Is the following the correct x264 mail list?
yes, it is. The libx264 guy pointed me to this link: http://git.videolan.org/gitweb.cgi?p=x264.git;a=commit;h=1adf1023fff7887df15... It looks it was a libx264 issue. Latest ffmpeg builds (zeranoe) include the fixed libx264 library and everything is fine now. Thank you very much to all! MB
Massimo Battistel <battistel <at> gmail.com> writes:
It looks it was a libx264 issue.
Latest ffmpeg builds (zeranoe) include the fixed libx264 library and everything is fine now.
Thank you for testing and reporting back! Carl Eugen
Massimo Battistel <battistel <at> gmail.com> writes:
2. Is there any relation between my issue and #993? ( http://trac.ffmpeg.org/ticket/993)
Only if you problem is not reproducible with -threads 1
The problem persists with -threads 1 (and ffmpeg command line).
Ticket #993 was a h264 decoding regression only reproducible with multi-threaded decoding, if I understand correctly you are reporting a problem with h264 encoding. Carl Eugen
participants (2)
-
Carl Eugen Hoyos -
Massimo Battistel