I have a source video that's 29.97 with 3:2 pulldown, and I want to encode it at 23.98p. This ffmpeg command produces video at 23.98, but with a pattern of 2 progressive frames followed by 2 interlaced frames. ffmpeg -ss 208 -i NCIS.ts -t 30 -vf "pullup,fps=24000/1001" -acodec copy -vcodec libx264 pullup.mkv On the other hand, I can achieve perfect results with this simple AviSynth script: MPEG2Source("NCIS.d2v", cpu=0) TFM().TDecimate() Trim(5000,5300) Since AviSynth isn't doing any deinterlacing, I don't think I should need to add yadif in ffmpeg. How can I make the pullup filter work correctly?
Elliott Balsley <elliottbalsley <at> gmail.com> writes:
This ffmpeg command produces video at 23.98, but with a pattern of 2 progressive frames followed by 2 interlaced frames.
ffmpeg -ss 208 -i NCIS.ts -t 30 -vf "pullup,fps=24000/1001" -acodec copy -vcodec libx264 pullup.mkv
(Complete, uncut console output missing.) Please provide the input sample. Note that I don't think you can combine pullup and a deinterlacer, this only works with fieldmatch iiuc. Carl Eugen
On Wed, Oct 02, 2013 at 11:28:56AM -0700, Elliott Balsley wrote: [...]
On the other hand, I can achieve perfect results with this simple AviSynth script:
MPEG2Source("NCIS.d2v", cpu=0) TFM().TDecimate() Trim(5000,5300)
TFM/TDecimate are ported to FFmpeg as fieldmatch/decimate; you may want to try. [...] -- Clément B.
Here's the full output: http://pastebin.com/kWmqgKRA I tried fieldmatch/decimate, but it made the audio out of sync. I reported it as a bug here: https://trac.ffmpeg.org/ticket/3019 I thought pullup was supposed to be the modern, better way of doing reverse pulldown. Here's an input sample. MPEG2 video with 5.1 AC3 audio. The header might be broken because I trimmed it using dd, but it still works. https://docs.google.com/file/d/0B52QuT8oHvtZcDNFNjdTaXpNdms/ Thanks for your help. On Oct 2, 2013, at 11:46 AM, Clément Bœsch <u@pkh.me> wrote:
On Wed, Oct 02, 2013 at 11:28:56AM -0700, Elliott Balsley wrote: [...]
On the other hand, I can achieve perfect results with this simple AviSynth script:
MPEG2Source("NCIS.d2v", cpu=0) TFM().TDecimate() Trim(5000,5300)
TFM/TDecimate are ported to FFmpeg as fieldmatch/decimate; you may want to try.
[...]
-- Clément B. _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
On 10/2/13, Elliott Balsley <elliottbalsley@gmail.com> wrote:
Here's the full output: http://pastebin.com/kWmqgKRA
I tried fieldmatch/decimate, but it made the audio out of sync. I reported it as a bug here: https://trac.ffmpeg.org/ticket/3019 I thought pullup was supposed to be the modern, better way of doing reverse pulldown.
It was silly bug in pullup filter. The latest master from git should not have this bug.
Elliott Balsley <elliottbalsley <at> gmail.com> writes:
I have a source video that's 29.97 with 3:2 pulldown, and I want to encode it at 23.98p. This ffmpeg command produces video at 23.98, but with a pattern of 2 progressive frames followed by 2 interlaced frames.
This bug was fixed by Paul.
ffmpeg -ss 208 -i NCIS.ts -t 30 -vf "pullup,fps=24000/1001" -acodec copy -vcodec libx264 pullup.mkv
Unfortunately, this command line still does not work: Instead of producing different frames, for every five (progressive and interlaced) input frames, four progressive output frames are produced, but only three different ones: The wrong frame is dropped;-( The following works fine here for your sample: ffmpeg -i input -vf pullup -r 24000/1001 out.mkv Carl Eugen
On 10/9/13, Carl Eugen Hoyos <cehoyos@ag.or.at> wrote:
Elliott Balsley <elliottbalsley <at> gmail.com> writes:
I have a source video that's 29.97 with 3:2 pulldown, and I want to encode it at 23.98p. This ffmpeg command produces video at 23.98, but with a pattern of 2 progressive frames followed by 2 interlaced frames.
This bug was fixed by Paul.
ffmpeg -ss 208 -i NCIS.ts -t 30 -vf "pullup,fps=24000/1001" -acodec copy -vcodec libx264 pullup.mkv
Unfortunately, this command line still does not work: Instead of producing different frames, for every five (progressive and interlaced) input frames, four progressive output frames are produced, but only three different ones: The wrong frame is dropped;-(
The following works fine here for your sample: ffmpeg -i input -vf pullup -r 24000/1001 out.mkv
Wouldn't using decimate after pullup be better approach?
Carl Eugen
_______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
On 10/9/13, Carl Eugen Hoyos <cehoyos@ag.or.at> wrote:
Paul B Mahol <onemda <at> gmail.com> writes:
The following works fine here for your sample: ffmpeg -i input -vf pullup -r 24000/1001 out.mkv
Wouldn't using decimate after pullup be better approach?
This does not really help with the original issue, don't you agree? ;-)
Indeed, thus pts issue(s) is decimate filter should be resolved.
Carl Eugen
_______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
Paul B Mahol <onemda <at> gmail.com> writes:
The following works fine here for your sample: ffmpeg -i input -vf pullup -r 24000/1001 out.mkv
Wouldn't using decimate after pullup be better approach?
This does not really help with the original issue, don't you agree? ;-)
Indeed, thus pts issue(s) is decimate filter should be resolved.
Carl Eugen
I'm now getting proper results with "-vf pullup -r 24000/1001". Are you saying decimate after pullup would be better? Why?
Elliott Balsley <elliottbalsley <at> gmail.com> writes:
I'm now getting proper results with "-vf pullup -r 24000/1001".
This is - afaict - the only variant that works atm.
Are you saying decimate after pullup would be better?
"-r" is generally deprecated (vf fps should be used), decimate is the filter to be used after the inverse telecine filters (but it does not work correctly atm). Carl Eugen
Sorry if this is now an old topic, but I am still facing exactly the same issue: pullup is always generating 2 progressive and 2 blended frames, whether I use vf "pullup, fps=24000/1001" or -vf pullup -r 24000/1001. I could submit sample and log, but I just would like to know if people think this issue has now been resolved satisfactorily in recent ffmpeg releases (I am using one I installed 2 days ago). -- View this message in context: http://ffmpeg-users.933282.n4.nabble.com/IVTC-with-pullup-filter-tp4661581p4... Sent from the FFmpeg-users mailing list archive at Nabble.com.
On 2/11/15, marcjn <windbreiz@gmail.com> wrote:
Sorry if this is now an old topic, but I am still facing exactly the same issue: pullup is always generating 2 progressive and 2 blended frames, whether I use vf "pullup, fps=24000/1001" or -vf pullup -r 24000/1001. I could submit sample and log, but I just would like to know if people think this issue has now been resolved satisfactorily in recent ffmpeg releases (I am using one I installed 2 days ago).
Did you tried fieldmatch & decimate filters?
-- View this message in context: http://ffmpeg-users.933282.n4.nabble.com/IVTC-with-pullup-filter-tp4661581p4... Sent from the FFmpeg-users mailing list archive at Nabble.com. _______________________________________________ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
marcjn <windbreiz <at> gmail.com> writes:
I could submit sample and log
If you don't submit them, help is extremely unlikely afaict. (This is of course not pullup related.) Carl Eugen
Carl Eugen Hoyos wrote
marcjn
gmail.com> writes:> I could submit sample and logIf you don't submit them, help is extremely unlikely afaict.(This is of course not pullup related.)Carl Eugen_______________________________________________ffmpeg-user mailing list
ffmpeg-user@
I don't see my recent mails to the list posted here so I will try from my browser (I am new to this mail list format).I have been trying to convert a telecined clip (29.97 fps) to a progressive one (23.976 fps).The clip I am using for test purpose is here: https://dl.dropboxusercontent.com/u/71351606/TermSample2.mp4 <https://dl.dropboxusercontent.com/u/71351606/TermSample2.mp4> The sample contains a x.264 video stream and an AAC 2-channel audio stream. I first tried to use -vf pullup, fps=24000/1001: The log is here <https://dl.dropboxusercontent.com/u/71351606/pullup_fps.log> The result was a 23.976 fps clip, showing up as 'progressive' with MediaInfo, but with 2 progressive frames, followed by 2 frames each blending 2 fields (combing). I also tried -vf pullup -r 24000/1001 as some users mentioned that this worked for them:Here is the log: here <https://dl.dropboxusercontent.com/u/71351606/pullup_rate.log> The result was the same: 2 progressive frames followed by 2 blended ones. I then tried -vf fieldmatch,decimate: The log is here <https://dl.dropboxusercontent.com/u/71351606/fieldmatch_decimate.log> Different result here: most of the time, we have 3 progressive + 1 blended frame, but on occasion 2 progressive + 2 blended.I do hope someone knows how to make IVCT work under ffmpeg.Thanks. -- View this message in context: http://ffmpeg-users.933282.n4.nabble.com/IVTC-with-pullup-filter-tp4661581p4... Sent from the FFmpeg-users mailing list archive at Nabble.com.
On Wednesday, February 11, 2015 1:34 PM, marcjn <windbreiz@gmail.com> wrote:
Carl Eugen Hoyos wrote marcjn
gmail.com> writes:> I could submit sample and logIf you don't submit them, help is extremely unlikely afaict.(This is of course not pullup related.)Carl Eugen_______________________________________________ffmpeg-user mailing list
ffmpeg-user@
I don't see my recent mails to the list posted here so I will try from my browser (I am new to this mail list format).I have been trying to convert a telecined clip (29.97 fps) to a progressive one (23.976 fps).The clip I am using for test purpose is here: https://dl.dropboxusercontent.com/u/71351606/TermSample2.mp4 <https://dl.dropboxusercontent.com/u/71351606/TermSample2.mp4> The sample contains a x.264 video stream and an AAC 2-channel audio stream. I first tried to use -vf pullup, fps=24000/1001: The log is here <https://dl.dropboxusercontent.com/u/71351606/pullup_fps.log> The result was a 23.976 fps clip, showing up as 'progressive' with MediaInfo, but with 2 progressive frames, followed by 2 frames each blending 2 fields (combing). I also tried -vf pullup -r 24000/1001 as some users mentioned that this worked for them:Here is the log: here <https://dl.dropboxusercontent.com/u/71351606/pullup_rate.log> The result was the same: 2 progressive frames followed by 2 blended ones. I then tried -vf fieldmatch,decimate: The log is here <https://dl.dropboxusercontent.com/u/71351606/fieldmatch_decimate.log> Different result here: most of the time, we have 3 progressive + 1 blended frame, but on occasion 2 progressive + 2 blended.I do hope someone knows how to make IVCT work under ffmpeg.Thanks.
I think your problem is that your media is not actually 29.97 fps. ffprobe reports 28.97 fps, multiplying that by 4/5 gives 23.76 Try just -vf pullup,fps=23.176 no fps argument. Worked here. -Nick
On Wednesday, February 11, 2015 3:21 PM, Nicholas Robbins <nickrobbins@yahoo.com> wrote:
I think your problem is that your media is not actually 29.97 fps.
ffprobe reports 28.97 fps, multiplying that by 4/5 gives 23.76
Try just -vf pullup,fps=23.176 no fps argument.
Don't know why I said "no fps argument". Try -vf pullup,fps=23.176 -Nick
On Wednesday, February 11, 2015 3:24 PM, Nicholas Robbins <nickrobbins@yahoo.com> wrote:
On Wednesday, February 11, 2015 3:21 PM, Nicholas Robbins <nickrobbins@yahoo.com> wrote:
I think your problem is that your media is not actually 29.97 fps.
ffprobe reports 28.97 fps, multiplying that by 4/5 gives 23.76
Try just -vf pullup,fps=23.176 no fps argument.
Don't know why I said "no fps argument".
Try -vf pullup,fps=23.176
Or try a different output format. If you use mkv as the output it seems to work, and counting frames gives a fps of 22.109 fps. Where are you getting this video? 28.97 fps is very strange. --Nick
Nicholas Robbins-2 wrote
On Wednesday, February 11, 2015 3:24 PM, Nicholas Robbins <
nickrobbins@
> wrote:
On Wednesday, February 11, 2015 3:21 PM, Nicholas Robbins <
nickrobbins@
> wrote:
I think your problem is that your media is not actually 29.97 fps.
ffprobe reports 28.97 fps, multiplying that by 4/5 gives 23.76
Try just -vf pullup,fps=23.176 no fps argument.
Don't know why I said "no fps argument".
Try -vf pullup,fps=23.176
Or try a different output format. If you use mkv as the output it seems to work, and counting frames gives a fps of 22.109 fps. Where are you getting this video? 28.97 fps is very strange.
--Nick _______________________________________________ ffmpeg-user mailing list
ffmpeg-user@
My mistake. I can't believe I did that and overlooked it . Thanks for looking at the sample. The reason for the strange frame rate: I created that clip from a 29.97 fps telecined recording, and to be sure I had a constant 29.97 fps, I wrote -r 29000/1001 instead of 30000/1001. I am adding a hopefully 'correct' 29.97 fps TermSample3.mp4 sample here <https://dl.dropboxusercontent.com/u/71351606/TermSample3.mp4> Now, using -vf pullup,fps=23.177 on the original sample still gives me 2 progressive + 2 blended frames when runnning the command (see log here <https://dl.dropboxusercontent.com/u/71351606/pullup_fps20.log> ) Using -vf pullup,fps=24000/1001 on the new 29.97 fps sample (TermSample3.mp4), I still get the same result (see log here <https://dl.dropboxusercontent.com/u/71351606/pullup_fps3.log> ) I will look at mkv as well but it seems that my frame rate error doesn't explain the issue at this point. -- View this message in context: http://ffmpeg-users.933282.n4.nabble.com/IVTC-with-pullup-filter-tp4661581p4... Sent from the FFmpeg-users mailing list archive at Nabble.com.
On Wednesday, February 11, 2015 4:40 PM, marcjn <windbreiz@gmail.com> wrote:
My mistake. I can't believe I did that and overlooked it . Thanks for looking at the sample. The reason for the strange frame rate: I created that clip from a 29.97 fps telecined recording, and to be sure I had a constant 29.97 fps, I wrote -r 29000/1001 instead of 30000/1001. I am adding a hopefully 'correct' 29.97 fps TermSample3.mp4 sample here <https://dl.dropboxusercontent.com/u/71351606/TermSample3.mp4> Now, using -vf pullup,fps=23.177 on the original sample still gives me 2 progressive + 2 blended frames when runnning the command (see log here <https://dl.dropboxusercontent.com/u/71351606/pullup_fps20.log> ) Using -vf pullup,fps=24000/1001 on the new 29.97 fps sample (TermSample3.mp4), I still get the same result (see log here <https://dl.dropboxusercontent.com/u/71351606/pullup_fps3.log> )
I will look at mkv as well but it seems that my frame rate error doesn't
explain the issue at this point. I can't reproduce the error. Leaving off the fps command, I get something that looks fine.
Using ffmpeg -i TermSample3.mp4 -vf pullup out.mkv I compare the frames in the original that were not combed, they appear to be identical to the new one. I don't see the problems you are seeing. The only problem is that this produces 24 fps material that claims to be 30 fps. Adding in -r 24000/1001 fixes the 30fps claim. Now the material claims to be 24fps. Only 2 frames are dropped as a result of the -r output option. -vf fieldmatch,decimateproduces material that looks identical to me. Could you double check that the orginal isn't blurred. Those amo belts or whatever he is swinging are blurred in the original because they are moving quickly. -Nick
Nicholas Robbins-2 wrote
Using
ffmpeg -i TermSample3.mp4 -vf pullup out.mkv
I compare the frames in the original that were not combed, they appear to be identical to the new one. I don't see the problems you are seeing. The only problem is that this produces 24 fps material that claims to be 30 fps.
Adding in -r 24000/1001 fixes the 30fps claim. Now the material claims to be 24fps. Only 2 frames are dropped as a result of the -r output option. -Nick _______________________________________________ ffmpeg-user mailing list
ffmpeg-user@
Now I see the same thing: Gives all progressive frames. Good result. Could'nt see the frame rate with MediaInfo, but has to be 23.976 fps All progressive frames: 4 individual frames + 1 duplicate. 29.97 fps. Just need to eliminate the dupes. All progressive frames, 23.976 fps, the dupes have been dropped. Good result. No idea why the behavior is different for the mp4 container and mkv. Also, I need to find out why works perfectly, while Nicholas Robbins-2 wrote
Could you double check that the orginal isn't blurred. Those amo belts or whatever he is swinging are blurred in the original because they are moving quickly.
The ammo belts are blurred because of the movement. Now, I need to find out why works well, while still creates 3 progressive + 1 blended. Jean -- View this message in context: http://ffmpeg-users.933282.n4.nabble.com/IVTC-with-pullup-filter-tp4661581p4... Sent from the FFmpeg-users mailing list archive at Nabble.com.
marcjn wrote
Now, I need to find out why
works well, while
still creates 3 progressive + 1 blended.
Jean
I found how to fix it: for some reason, the crop filter before pullup works OK, but the scale filter prevents pullup from working well. So the right way is to put the pullup filter first. This was the reason why I couldn't make it work all that time. Now, this works: I think my problem is solved now. Thank you Nick. -- View this message in context: http://ffmpeg-users.933282.n4.nabble.com/IVTC-with-pullup-filter-tp4661581p4... Sent from the FFmpeg-users mailing list archive at Nabble.com.
participants (6)
-
Carl Eugen Hoyos -
Clément Bœsch -
Elliott Balsley -
marcjn -
Nicholas Robbins -
Paul B Mahol