Re: [FFmpeg-user] Deinterlacing detected interlaced frames
On 02/17/2015 01:12 PM, Carl Eugen Hoyos wrote: Iirc, the mailing list rules ask you not to compress input if not necessary (maybe I misremember). Yes, you're right, I'm terribly sorry I forgot to read them: I followed general netiquette and netiquette from Gentoo mailing lists and didn't want to make anyone feel flooded (and wanted to lower mailing lists' overhead). In any case, if you convert a 42 frame input sample, there is no need to attach something that people have to download (and uncompress), just inline the console output. Thank you for saying nicely, I'll remember it when sending another log. I don't think idet detects the single interlaced frame. I think it can (according to the log) but I am not sure how the compress phase handles it later. I made a script which takes the detected positions of interlaced frames and merges timecodes which are near (assuming ffmpeg doesn't detect every frame) and it works quite well but I'd still like to use directly ffmpeg for it. -- Jan Sever ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
Re: [FFmpeg-user] Deinterlacing detected interlaced frames
On 02/15/2015 04:54 PM, Carl Eugen Hoyos wrote: Please provide the complete, uncut console output (internal encoder preferred) and the input sample. I did in previous post but maybe you mean debug loglevel, so I ran: ffmpeg -loglevel repeat+debug -i in.mkv -vf idet,yadif=deint=interlaced \ out.mkv out.log out.log.xz is in the attachment. To get an input sample: wget -qO- 'http://pastebin.com/download.php?i=En8vREKf' | tr -d \\r \ | base64 -d in.mkv In my sample only frame 11 is interlaced but ffmpeg doesn't deinterlace it, although it detects interlacing. Thanks in advance. -- Jan Sever out.log.xz Description: application/xz ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
[FFmpeg-user] Deinterlacing detected interlaced frames
Hi all, I'm trying to deinterlace badly deinterlaced videos (most frames progressive, some interlaced) with ffmpeg -i in.mkv -vf idet,yadif=deint=interlaced -c:v libx264 -an -y out.mkv but deinterlacing looks like shifted (it deinterlaces another frames than marked). I looked at output of ffmpeg -loglevel debug -i in.mkv -vf idet -an -f rawvideo -y /dev/null and it detects the interlaced frames correctly. Is there any problem with deint=interlaced or do I need to add another parameter? Log of the first command attached. Thank you in advance for your precious help. -- Jan Sever out.log.xz Description: application/xz ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
[FFmpeg-user] Finding which frames were marked as interlaced by vf=idet
Hi all, is there any option to enable printing of numbers of frames which were marked as interlaced by vf=idet? I'd like to detect the parts in a movie that are interlaced because the detection is not 100% successfull (although it's very precise: many thanks to everybody involved), so I'd like to detect interlaced frames and then if there is a bunch of interlaced frames, although not subsequent, to deinterlace from the first to the last, deinterlacing even a few frames before the first and after the last detected. Thanks in advance for your precious advice, Jan Sever ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
[FFmpeg-user] Applying filters for selected frames?
Hi all, is it possible to apply filters (-vf) in ffmpeg only for selected frames? E. g. to apply a filter for first ten minutes and not apply for the rest of movie? I'm now solving a problem of badly deinterlaced movies, which have some parts not deinterlaced while the rest is. I didn't find this possibility in man page so I assume it's not supported (yet). Is it possible to add the support for example in this form: ffmpeg ... -vf yadif=0-10:00,20:00-30:00 ... to get only first and third 10 minutes deinterlaced? Thank you in advance, Jan Sever ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
Re: [FFmpeg-user] Applying filters for selected frames? [SOLVED]
On 12/21/2014 8:53 PM, Moritz Barsnick wrote: My bad, thanks for the clarification. You needn't apologize at all, your help is very notable. I think I had this thread in the back of my mind: http://ffmpeg.org/pipermail/ffmpeg-user/2014-May/021412.html which introduces use of the idet filter to detect interlacing (type). Wow, I didn't know about this: it solves my problem even much more better. So Jan (the OP) will have to check what applies to his material. Timeline-editing is exactly what I asked for, but you wrote me about even better solution (previous note). So in other words both solutions are excellent but the first is better ;-), marking as SOLVED. Many thanks to everybody, you solved the problem faster than very quickly. Jan Sever P.S. Sorry for keeping you waiting so long for my answer, I had to unmask newer version of ffmpeg first (in Gentoo ffmpeg-2 is still marked as unstable), recompile it and test both solutions (I didn't have a usable mixed record with both progressive and interlaced frames /big compression prevented from detection of interlacing/, so I made it from two). ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
Re: [FFmpeg-user] Applying filters for selected frames? [SOLVED]
On 12/21/2014 11:16 PM, Nicholas Robbins wrote: That was my original thread. I've worked through various combination of interlaced, progressive, telecined, filmrate, etc. I've worked out various ffmpeg settings that work for most of these situations. Post if you have one that you are not happy with (for mixed progressive interlaced, I use -vf idet,yadif=mode=1:deint=interlaced,fps=fps=6/1001 and that has worked. You can remove the fps filter at the end if you don't care about producing CFR video. Thank you for this thread, I used the parameters (save fps) and it worked like a charm: I had been sad of mixed records from Czech Television but now I can see there's nothing Open Source would be unable of. ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
Re: [FFmpeg-user] Worse quality than mencoder [SOLVED]
On 2014-11-26 12:30, Moritz Barsnick wrote: Are you absolutely sure? I'm sorry, they're the same only for the second pass: that was the problem. You're doing two-pass, right? If that is why these logs each contain two outputs from the encoders, then you're doing something significantly wrong: mencoder.txt:x264 [info]: profile Main, level 3.0 mencoder.txt:x264 [info]: profile High, level 4.0 ffmpeg.txt:[libx264 @ 0x63c9c0] profile Main, level 4.0 ffmpeg.txt:[libx264 @ 0x63c9c0] profile High, level 4.0 Level was not the problem. The libx264 settings from the first pass and second pass differ significantly from each other - and in a different manner in ffmpeg and mencoder. Only the second pass settings seem identical, which may be why you see the same embedded in the file. Yes, you're right, that was the problem. The encoders set some parameters for quicker first pass (I knew that), but what I didn't know that ffmpeg doesn't set fast_pskip=1 and ref=1 while mencoder does. AFAIU you need to use identical codec options in both passes, and it seems you're not doing that. That _may_ be the cause for your observations. In fact, it is. I compared the achieved quality after changing the two mentioned parameters and it looks almost the same now. So marking as SOLVED. Thanks everybody for your precious help. Jan Sever ___ ffmpeg-user mailing list ffmpeg-user@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-user
