[Ffmpeg-devel-irc] ffmpeg.log.20151022

burek burek021 at gmail.com
Fri Oct 23 02:05:01 CEST 2015


[00:00:37 CEST] <onefix> You could build a script that could be ran as a daemon that watches for a directory to show up and starts encoding a part
[00:01:07 CEST] <onefix> If you did it right, the main system wouldn't need to know anything about the number of encoders available, it could just deposit the files and wait for them to finish
[01:12:21 CEST] <BullHorn> erm
[01:12:43 CEST] <BullHorn> ive been using my PC for a couple years, mostly gaming+recording video, never issues and the CPU never went above ~70c or so
[01:12:53 CEST] <BullHorn> im now re-encoding a video with ffmpeg for the last 2 hours and http://i.imgur.com/XM7j8AH.png
[01:12:57 CEST] <BullHorn> a little spoopy D:
[01:17:17 CEST] <BullHorn> is there a way to use ffmpeg without avx2
[01:17:24 CEST] <dsl420> when i do encode on a notebook it looks like that, for an intel it might be Okish :)
[01:17:31 CEST] <BullHorn> it raises my CPU's voltage beyond the limit i set it to
[01:18:16 CEST] <dsl420> well 90°C is pretty high i guess i encode at 85°C on a notebook
[01:20:16 CEST] <drv> 100C is usually the thermal limit where throttling starts to happen
[01:20:23 CEST] <drv> looks like insufficient/incorrectly installed cooling
[01:20:57 CEST] <BullHorn> it never went above 70
[01:21:04 CEST] <BullHorn> on the other hand i never used ffmpeg for 2h straight
[01:21:15 CEST] <BullHorn> maybe its time to clean the dust filters and re-apply thermal paste
[01:21:25 CEST] <BullHorn> ill do it during the weekend and avoid ffmpeg until then ;x
[01:22:30 CEST] <BullHorn> i hope 2 year old thermal paste tube is still good
[02:26:11 CEST] <BullHorn> hmm i need some help
[02:26:16 CEST] <BullHorn> i currently use this command line:
[02:26:18 CEST] <BullHorn> ffmpeg -i D:\Streaming\%1.mkv -c:a copy -c:v libx264 -crf 24 -preset:v veryfast -copyts F:\%2.mp4
[02:26:33 CEST] <BullHorn> issue is: i started recording 2 audio tracks but this only copies the first track
[02:27:38 CEST] <xsi> [ALSA] Every audio codec which I use with ffmpeg has intermittent sound. I changed from libmp3lame to libfaac to aac to adpcm_swf (all compatible with .flv, for twitch screencast). The same intermittent sound.
[02:28:11 CEST] <xsi> What could it be with the sound recording? (ffmpeg): https://youtu.be/XNT0pSHb5oI . My command line is: https://bpaste.net/show/ee4c3b03a95e
[02:29:04 CEST] <xsi> sorry maybe i'm annoying #ffmpeg channel but my sound sucks(
[02:31:06 CEST] <furq> BullHorn: afaik you have to use -map
[02:31:14 CEST] <furq> probably -map 0:0 -map 0:1 -map 0:2
[02:32:13 CEST] <BullHorn> can you help me modify my line with this?
[02:32:21 CEST] <BullHorn> i keep the line in a .bat file so i just convert a b and it goes
[02:32:32 CEST] <BullHorn> although i have no idea about bat files or ffmpeg ;x
[02:32:39 CEST] <furq> put it after -c:a copy
[02:32:55 CEST] <furq> anywhere between the two filenames is fine
[02:33:06 CEST] <xsi> [ALSA] Every audio codec which I use with ffmpeg has intermittent sound. I changed from libmp3lame to libfaac to aac to adpcm_swf (all compatible with .flv, for twitch screencast). The same intermittent sound when recording - what could it be? : (ffmpeg): https://youtu.be/XNT0pSHb5oI . My
[02:35:26 CEST] <BullHorn> thats how it looks now:
[02:35:26 CEST] <BullHorn> ffmpeg -i D:\Streaming\%1.mkv -map 0 -c:a copy -c:v libx264 -crf 23 -preset:v veryfast -copyts F:\%2.mp4
[02:35:30 CEST] <BullHorn> ill go try
[02:35:42 CEST] <furq> that's totally not what i said
[02:36:08 CEST] <BullHorn> i thought -map 0 would take all the values including :0, :1, :2 etc
[02:36:08 CEST] <BullHorn> ;x
[02:36:15 CEST] <furq> i do approve of the crf mysterously increasing though
[02:36:35 CEST] <BullHorn> a 76gb file recorded with ultrafast went down to 30gb with veryfasy
[02:36:43 CEST] <furq> oh that is actually right
[02:36:44 CEST] <BullHorn> i can handle a bit more filesize/quality :p
[02:36:46 CEST] <furq> the -map 0 thing
[02:37:17 CEST] <furq> i've only ever used -map to remove russian audio tracks
[02:37:24 CEST] <BullHorn> :D
[02:37:29 CEST] <BullHorn> cyka blaty
[02:37:34 CEST] <furq> sorry gavrilov
[02:37:42 CEST] <BullHorn> drink dotka play vodka
[02:39:54 CEST] <BullHorn> yes this worked furq
[02:39:58 CEST] <BullHorn> -map 0 was all it took :)
[06:37:05 CEST] <ohsnap> hey all. im running ffmpeg with the command line options " -i blahblah.mkv -c copy out.mp4 " to break a video out of a mkv container, but im a noob and i cant figure out how to change the audio track it dumps out with it
[06:37:16 CEST] <ohsnap> this container had several different audio tracks available and i need a different one
[06:55:44 CEST] <ohsnap> so yeah i rtfmd. -map
[06:55:50 CEST] <ohsnap> stay seksi <3
[08:18:48 CEST] <deetwelve> Hello, I am using ffmpeg and rtmp to stream a file. Is there a command to start the file at a specific time?
[08:27:24 CEST] <wt> hi...is is possible to pass args via a file instead of via the commandline?
[12:16:09 CEST] <maurobaraldi> Hello world. Does the ffmpeg load entire file when read or there is a way to do so?
[13:58:35 CEST] <pgunnars> where can i see the log for this channel
[14:02:25 CEST] <waressearcher2> is that channel logged ?
[14:06:57 CEST] <c_14> http://ffmpeg.gusari.org/irclogs/
[14:48:49 CEST] <pgunnars> BtbN: you here?
[14:55:50 CEST] <BtbN> Not for too long
[14:55:58 CEST] <BtbN> like, 5 minutes or so
[14:59:58 CEST] <BtbN> I do read the backlog though, so if you're fine with a few hours of delay...
[15:06:59 CEST] <pgunnars> was just wondering regarding the questions i was asking you yesterday
[15:07:25 CEST] <pgunnars> the difference between outputting to "-" compared to "file.webm"
[15:07:58 CEST] <BtbN> hm?
[15:08:00 CEST] <pgunnars> lose the metadata when I create a file from the stdout output
[15:08:13 CEST] <pgunnars> theres no duration etc
[15:08:40 CEST] <pgunnars> how to add the duration info to the file
[15:11:25 CEST] <BtbN> by using a seekable output format.
[15:12:40 CEST] <pgunnars> could you elaborate?
[15:13:46 CEST] <pgunnars> well, I don't think that's really viable for the application, I'm sending the request directly to a http request
[15:13:57 CEST] <pgunnars> Don't want to have to extract it to a file and then send the file
[15:15:38 CEST] <pgunnars> the stream directly..*
[16:20:48 CEST] <tonsofpcs> I have an hour-long .ts, I want to extract the last 5 minutes of one of the ESs (video, MPEG2, 1920x1080, 30000/1001 frames per second, interlace), and possibly decode it too - what's the best way to do this? (I don't have the HDD space to decide the entire ES yet alone the entire file)
[16:21:19 CEST] <tonsofpcs> PCR-on-Video, yes, I know what all of the PIDs are
[16:28:09 CEST] <pgunnars> BtbN: will just transcode server side I guess, will probably need to use ffmpeg there anyway
[16:30:32 CEST] <TikityTik> Why doesn't -to work?
[16:33:19 CEST] <DHE> if it's CBR you can cut the file with 'dd'. it sometimes causes minor weirdness at the start of the video but it works
[16:34:18 CEST] <TikityTik> I want it to be duration based though
[16:35:49 CEST] <TikityTik> i got it working nvm
[16:35:53 CEST] <BtbN> pgunnars, ffmpeg can't seek back to the beginning of the file if you are writing it to a pipe.
[16:36:05 CEST] <BtbN> But it needs to do that, so it can write the information it only knows at the end. Like the duration.
[16:36:44 CEST] <BtbN> You don't have to do full transcoding on the server. Just do a full -c copy of the file.
[16:37:05 CEST] <pgunnars> cool
[16:38:22 CEST] <pgunnars> thanks
[17:39:36 CEST] <buu> This is a vague and slightly hard to research question, but is there any way of asking ffmpeg or another tool if a video can be compressed without losing quality?
[17:45:34 CEST] <flux> buu, it seems to me ffmpeg does not store such an information in a general fashion about the codecs. so, no.
[17:47:29 CEST] <JEEB> lol
[17:47:36 CEST] <JEEB> use lossless compression, receive bacon
[17:47:42 CEST] <JEEB> lossy compression *always* loses quality
[18:18:00 CEST] <RobotsOnDrugs> also receive gigantic file sizes
[18:18:30 CEST] <RobotsOnDrugs> but less gigantic than uncompressed, of course
[18:37:30 CEST] <buu> I meant without a major loss in quality =/
[18:37:34 CEST] <buu> Not literally zero
[18:38:07 CEST] <buu> Although I am downloading youtube videos so presumably google has a pretty major interest in compressing them
[19:05:26 CEST] <RobotsOnDrugs> if you're downloading from youtube, why not just leave them the way they are?
[19:51:29 CEST] <TD-Linux> buu, well there are metrics you can use like SSIM to measure the loss in quality. However, youtube already uses a constant quality metric, so they've already done the "analysis" for you
[19:52:30 CEST] <TD-Linux> if the files are too big, make sure youtube-dl is grabbing the VP9 version. and then if they are still too big, reduce your max resolution
[19:52:41 CEST] <buu> TD-Linux: Yeah, that ocurred to me afterwards
[19:53:03 CEST] <buu> Any analysis I'm going to do is unlikely to beat theirs... I hope
[20:15:35 CEST] <bofh_> Hi there! I have the following rather complex command to merge two files: http://pastebin.com/9JK0k5g8, but for some reason ffmpeg complains that it can't find stream from filter expression. I assume that it gets confused because those are the video files so I need to instruct ffmpeg to extract audio streams first - but not sure how to do that.
[20:15:40 CEST] <bofh_> please advice
[20:17:52 CEST] <bofh_> llogan: http://pastebin.com/1ryxfNRY
[20:23:49 CEST] <llogan> bofh_: works for me
[20:24:12 CEST] <llogan> oh, i see.
[20:24:34 CEST] <llogan> use [0:a] instead of [0]. and [1:a] instead of [1]
[20:24:51 CEST] <llogan> i just tested on audio only inputs
[20:26:32 CEST] <bofh_> interesting, let me try
[20:26:54 CEST] <llogan> ...and that's why your original command worked for me
[20:29:16 CEST] <bofh> llogan: okay, I think now I know the problem - one of the files doesn't contain an audio stream
[20:29:44 CEST] <bofh> is there any way to "ignore" it while applying filters? I may not know in advance which of input files may not contain an audio stream
[20:30:01 CEST] <llogan> oh, heh. that will do it too. didn't even notice.
[20:31:35 CEST] <llogan> as for ignoring; i doubt it but i'm not 100% sure.
[20:31:59 CEST] <llogan> you could use ffprobe and scripting
[20:32:34 CEST] <bofh> llogan: http://pastebin.com/m17kuhrE
[20:33:12 CEST] <llogan> your first input has no audio
[20:34:56 CEST] <bofh> yes, so perhaps I need to pre-filter the input data before starting the conversion
[20:49:48 CEST] <hero_biz> hi guys
[20:49:53 CEST] <hero_biz> a question.
[20:50:28 CEST] <hero_biz> how i can encode some parts of video with ffmpeg?
[20:51:19 CEST] <hero_biz> for example I want to only encode time 0 to 30 and 50 to 150 from a video.(time in seconds)
[20:54:14 CEST] <NitLord> hero_biz: each segment as separate outputs, or both concatenated into one?
[20:54:37 CEST] <hero_biz> no,for example original is a 200 sec video
[20:54:56 CEST] <hero_biz> and i want to encode those 2 part of original into outputfile
[20:55:31 CEST] <NitLord> so the 2nd option
[20:55:57 CEST] <hero_biz> ?? 2nd option?
[20:56:17 CEST] <NitLord> both segments concatenated into one output
[20:57:06 CEST] <hero_biz> yea,I want it like this
[20:57:13 CEST] <hero_biz> but how i should do it?
[20:57:33 CEST] <hero_biz> how I should specify those segments for encoding from original file?
[20:57:34 CEST] <NitLord> do you want different video/audio/container formats than the original?
[20:58:20 CEST] <hero_biz> i will be probably forced to encode it to another format.
[21:01:00 CEST] <NitLord> then use the concat filter. ffmpeg -t 30 -i input0 -ss 50 -t 100 -i input2 -filter_complex "[0:v][0:a][1:v][1:a]concat=n=2:v=1:a=1[v][a]" -map "[v]" -map "[a]" output
[21:01:28 CEST] <NitLord> or use the trim/atrim filters instead of -ss/-t
[21:02:17 CEST] <NitLord> alternatively, if you don't want to re-encode, you could make stream copied segments then use concat demuxer
[21:03:25 CEST] <hero_biz> ty :) I think atrim filter will solve my problem,if I could use it multiple times :)
[21:09:48 CEST] <hero_biz> can I use trim filter like this: -vf trim=0:30,50:100 ?
[21:10:15 CEST] <NitLord> no
[21:11:09 CEST] <NitLord> make a new trim instance per segment. also, don't forget that trim is just for video. atrim for audio.
[21:11:48 CEST] <hero_biz> but I have 1 input file...
[21:12:09 CEST] <hero_biz> which is 200 second long.
[21:12:59 CEST] <hero_biz> trim manual says: "If multiple start or end options are set, this filter tries to be greedy and keep all the frames that match at least one of the specified constraints. "
[21:13:14 CEST] <hero_biz> but I don't get what this means...
[21:13:17 CEST] <NitLord> ffmpeg -i input -filter_complex "[0:v]trim=0:30[t0];[0:v]trim=50:100[t1];[t0][t1]concat=n=2:v=1:a=0[v]" -map "[v]"
[21:14:19 CEST] <NitLord> i don't quite understand that sentence either.
[21:14:28 CEST] <hero_biz> O.O complex....
[21:15:27 CEST] <hero_biz> what happens to audio with this command that you have posted?
[21:15:33 CEST] <NitLord> it is ignored
[21:16:41 CEST] <NitLord> if you want it add some atrim. then change the concat filter, then map it
[21:19:02 CEST] <hero_biz> command will be like this? ffmpeg -i input -filter_complex "[0:v]trim=0:30[t0];[0:v]trim=50:100[t1];[0:a]atrim=0:30[a0];[0:a]atrim=50:100[a1];[t0][t1]concat=n=2:v=1:a=0[v];[a0][a1]concat=n=2:v=0:a=1[a]" -map "[v]" -map "[a]"
[21:20:04 CEST] <NitLord> almost. [t0][a0][t1][a1]concat=n=2:v=1:a=1[v][a]
[21:20:31 CEST] <hero_biz> got it
[21:20:56 CEST] <hero_biz> i wonder if audio and video will remain synched if I do like this....
[21:21:08 CEST] <NitLord> you'll just have to watch it
[21:21:17 CEST] <hero_biz> lol
[21:22:16 CEST] <hero_biz> maybe encoding whole video and then spliting it will be better...
[21:22:31 CEST] <hero_biz> split/concat
[21:28:01 CEST] <hero_biz> btw another question
[21:28:32 CEST] <hero_biz> i'm trying to deinterlace a video and have used yadif=1 filter.
[21:28:45 CEST] <hero_biz> original was 1080i.
[21:29:19 CEST] <hero_biz> my result video is 60 fps. how I go to 24000/1001 fps for output file?
[21:32:12 CEST] <NitLord> -r 24000/1001 or -r ntsc-film. or try the fps filter. the results may differ, but i can't remember what the difference is between them
[21:33:12 CEST] <NitLord> note that yadif=1 doubles the frame rate
[21:37:36 CEST] <hero_biz> if i use -vf yadif=1 -r 24000/1001, result will be 24000/1001, isn't it?
[21:38:04 CEST] <llogan> yes
[21:38:17 CEST] <llogan> why do you want to change frame rate?
[21:38:51 CEST] <hero_biz> it is 60fps... O.O I thought to change to common 24000/1001.
[21:39:19 CEST] <worstje> I currently pipe a raw file to x264.exe for rendering like this: ffmpeg.exe -i "raw.avi" -an -c:v wrapped_avframe -f yuv4mpegpipe -an -pix_fmt yuv420p - | x264.exe ...  but now I would like to squeeze in a small generated piece of footage at the start. Is there a way I can squeeze this extra footage in beforehand without having to re-encode my raw footage first?
[21:41:05 CEST] <furq> hero_biz: yadif=0 will output 30fps video
[21:41:17 CEST] <hero_biz> will decreasing output fps affect output file size too? (I think it should, just want to make sure)
[21:41:23 CEST] <furq> you probably want to stick with that rather than dropping random frames
[21:41:37 CEST] <furq> and yes it'll reduce the filesize
[21:41:38 CEST] <llogan> hero_biz: i would probably leave it at 60. all the frame dropping might look crappy (but it will depend on content).
[21:42:19 CEST] <llogan> you'll also have to compare yadif modes. sometimes 1 looks much better to me (again, all depending on the input)
[21:42:39 CEST] <furq> generally for film content yadif=1 looks wrong
[21:42:43 CEST] <hero_biz> when you deinterlace a 1080i video and scale it to 720p, what command line you normally use?
[21:43:03 CEST] <hero_biz> i chcked for animation,yadif=1 was good
[21:43:34 CEST] <furq> most stuff looks better with yadif=1
[21:43:43 CEST] <llogan> yadif,scale=-1:720
[21:43:50 CEST] <furq> you'll pay for it with filesize though
[21:44:49 CEST] <llogan> worstje: i don't really understand your question. also, why not just use libx264 instead of piping?
[21:45:03 CEST] <hero_biz> can we decerase frame rate in libx264 encoding process?
[21:45:23 CEST] <furq> -r will do that
[21:45:29 CEST] <furq> it won't look good though
[21:45:54 CEST] <hero_biz> -r drops frames randomly?
[21:45:57 CEST] <worstje> llogan: Because this is the way the script was before it ended up in my hands, and why break something that works?
[21:46:01 CEST] <furq> pretty much
[21:46:55 CEST] <hero_biz> I thought we can keep more important frame and drop less important ones when encoding and decreasing frame rate.
[21:47:04 CEST] <furq> which frames are less important
[21:47:28 CEST] <llogan> worstje: seems unecessarily complex and harder to maintain
[21:48:20 CEST] <hero_biz> @furq those which has lease change in frames.
[21:48:29 CEST] <hero_biz> *least
[21:48:37 CEST] <llogan> worstje: and what is "wrapped_avframe"?
[21:49:16 CEST] <c_14> llogan: the format used by the yuv4mpegpipe muxer
[21:49:26 CEST] <c_14> s/format/codec/
[21:49:28 CEST] <c_14> "codec"
[21:50:25 CEST] <furq> hero_biz: i'm not aware of any way to do that but maybe i've missed something
[21:50:30 CEST] <furq> -r certainly won't do that
[21:51:33 CEST] <furq> personally i would just use yadif=0 and leave it at 30fps
[21:51:44 CEST] <llogan> c_14: ah, it's new. less than 2 weeks old
[21:51:57 CEST] <hero_biz> then people that rip a ts interlaced video and you normally always see 24000/1001 fps, they have just used -r to decerase fps?
[21:52:31 CEST] <furq> i assume they've IVTC'd it
[21:53:43 CEST] <hero_biz> what is that?
[21:54:07 CEST] <furq> https://en.wikipedia.org/wiki/Telecine#Reverse_telecine_.28a.k.a._inverse_telecine_.28IVTC.29.2C_reverse_pulldown.29
[21:55:28 CEST] <furq> it only makes sense for film content which has duplicated frames at 30fps
[21:55:49 CEST] <furq> i live in a PAL region so i've never had much cause to use it
[22:12:11 CEST] <worstje> Why is the libx264 encoder creating too much footage? I am expecting it to match the sound file (exactly 5s), but I am getting a bit more than 7.2s whenever I try to create a file using this cmdline: ffmpeg.exe -loop 1 -y -i image.png -i sound.wav -c:a libfdk_aac -shortest -vcodec libx264 -pix_fmt yuv420p result.avi
[22:12:33 CEST] <worstje> Other encoders do in fact produce 5 seconds worth of footage.
[22:14:40 CEST] <worstje> Command output is here: http://pastebin.com/7gZtSg3L
[22:34:53 CEST] <muxan> hi i have a question about encoding into lossless(FLAC) specificly if the verification was ever implemented?
[22:36:32 CEST] <deetwelve> Hello, I am using ffmpeg and rtmp to stream a file. Is there a command to start the file at a specific time?
[22:43:35 CEST] <grublet> deetwelve: i dont know if it works for rtmp but you might want to try using -ss before the url
[22:44:02 CEST] <grublet> if that doesn't work maybe someone else can help you
[22:50:48 CEST] <deetwelve> i did try -ss with no luck
[22:50:55 CEST] <deetwelve> thanks though
[22:55:38 CEST] <onefix> So, anyone know if there are any scripts available to do split encoding using (for instance) and NFS share?
[22:59:51 CEST] <onefix> I have a script built for TVHeadend to re-encode  programs recorded from cable ... I would like to create an NFS share that I could point something like a Raspberry Pi to, it would determine the new encoding process and begin encoding 1 of many split files using a command of my determination
[23:01:10 CEST] <buu> Your rpi is going to encode hd video?
[23:08:21 CEST] <onefix> buu: Actually, a CHIP http://www.cnx-software.com/2015/05/08/chip-is-a-9-linux-development-board-powered-by-allwinner-r8-crowdfunding/
[23:09:36 CEST] <buu> What kind of fps do you get on that?
[23:09:37 CEST] <muxan> alwinner bleh
[23:10:26 CEST] <onefix> buu: It's a matter of scalability ... it may take 2 minutes to encode a 1 minute video on a CHIP, versus 30 seconds on a 6 core AMD ... but at $9 a piece, you could throw 10 of those at it for less than $100
[23:10:56 CEST] <buu> um.
[23:11:10 CEST] <buu> Those numbers sound unlikely.
[23:11:18 CEST] <muxan> totally
[23:11:19 CEST] <buu> But it sounds hilarious, so good luck
[23:11:33 CEST] <buu> Are you trying to chop a video into pieces?
[23:11:44 CEST] <onefix> buu: How so?  What kind of encoding times are we talking about?
[23:12:59 CEST] <onefix> buu: Right, I envision splitting say a 15 minute video into 100 pieces and then letting the "cluster" encode ... then waiting for all of them to complete and merging them back together
[23:13:07 CEST] <muxan> does ffmpeg use the official flac.exe to encode?
[23:14:26 CEST] <buu> onefix: I'd guess a $100 amd would encode at about 8 times the speed of an rpi, but maybe there's fpga magic I don't know about
[23:14:34 CEST] <onefix> buu: I gave up on hardware encoding a long time ago
[23:14:54 CEST] <buu> But I adore the idea of clusters of pis
[23:15:04 CEST] <DHE> ffmpeg doesn't use external apps typically. it would bundle a FLAC codec library
[23:15:21 CEST] <onefix> buu: Yeah, but if the rPi was $9, (like the CHIP), you could run 10 of them for the same $100
[23:15:42 CEST] <buu> Yes, but you aren't going to run 10 of them for $100
[23:15:53 CEST] <buu> Not that you get the cpu for just 100, but anyway
[23:16:00 CEST] <buu> onefix: The simplest case is that you're recording the cable show in a container that's splittable by a dumb program
[23:16:11 CEST] <onefix> muxan: IT would likely use the same library that FLAC.exe uses, just like it uses libx264 for h.264 encoding
[23:16:14 CEST] <muxan> i like to verify the encoded flac to the original. flac seems to use the -y flag for that
[23:16:27 CEST] <onefix> buu: It is, it's MPEG2 TS
[23:16:30 CEST] <buu> I think stuff like mpeg1/2 has very little look-ahead in the encoding
[23:16:32 CEST] <muxan> is there something simular in ffmpeg?
[23:16:54 CEST] <onefix> buu: It's coming directly from a HDHomerun
[23:17:02 CEST] <muxan> onefiz, ah ok
[23:17:34 CEST] <buu> hrm
[23:17:55 CEST] <llogan> worstje: the paste link has been removed
[23:18:25 CEST] <muxan> using the same lib doesn't always mean same features unfortunatly
[23:18:44 CEST] <buu> Ok, mpeg2 is actually way more complicated than I realized
[23:18:50 CEST] <onefix> buu: There is a script that processes the programs ... it runs COMSKIP to cut out commercials (if applicable), extracts closed captioning to a SRT file, and then reencodes to x264 + AAC and optionally resizes the video
[23:19:18 CEST] <buu> onefix: So why not just have that encode the files?
[23:19:30 CEST] <onefix> buu: There's also a test for cropping 480p video
[23:19:32 CEST] <furq> onefix: for fun i just started a 1080p > 576p x264 encode on my rpi
[23:19:51 CEST] <furq> i get about 65fps on an i7 920, on the rpi it's running at 2fps
[23:19:55 CEST] <buu> haha
[23:20:00 CEST] <furq> on an rpi2 clocked at 1.1ghz
[23:20:10 CEST] <buu> Oh wait, you can't hardware encode
[23:20:17 CEST] <buu> yeah that's just never going to go well
[23:20:29 CEST] <furq> idk if it could use hardware but i'm not doing
[23:20:48 CEST] <onefix> buu: It does, but I would like to see what I could make it do if all of my devices could start an encoding job ... presumably you could use Linux desktops and even Windows desktops to do the encoding if used CygWin
[23:20:51 CEST] <buu> furq: I had a conversation about it in here a while ago, rpi and related things can have decoder chips but not encoder chips
[23:21:01 CEST] <buu> onefix: dist-ffmpe
[23:21:04 CEST] <buu> what fun
[23:21:06 CEST] <furq> i don't have the decoders or anything installed anyway, i'm running it headless
[23:21:29 CEST] <furq> but yeah it's an order of magnitude slower than a core 2 duo at stock clocks
[23:21:30 CEST] <buu> furq: rpi2s come with a hardware x264 decocder, that's why they're able to play hd streams
[23:21:54 CEST] <furq> i should point out it's reading from and writing to a cifs share, because obviously it's going to be even slower writing to sd
[23:21:59 CEST] <furq> but i doubt that's making any difference at 2fps
[23:22:14 CEST] <buu> That's a lot of rpis for your cluster.
[23:22:23 CEST] <buu> furq: But just imagine if you had 80 of them.
[23:22:28 CEST] <buu> It'd go like lightning.
[23:22:45 CEST] <furq> you could probably do some pretty cool stuff with 80 rpis
[23:22:49 CEST] <furq> but i don't think video encoding is one of them
[23:22:57 CEST] <onefix> Yeah, the current box that I have doing the transcoding is a 6 core AMD ... I have built in use of a sqlite database and 2 /tmp files to do simultaneous encoding
[23:23:10 CEST] <muxan> 80x2 160fps
[23:23:13 CEST] <buu> simultaneous?
[23:23:30 CEST] <buu> You know ffmpeg is threaded right
[23:23:47 CEST] <furq> yeah x264 makes good use of multiple cores
[23:23:48 CEST] <muxan> rpi 30$x80 is $2400
[23:24:03 CEST] <furq> muxan: don't forget the power adapters
[23:24:09 CEST] <furq> unless you have a pc with 80 usb ports
[23:24:11 CEST] <muxan> oops
[23:24:35 CEST] <muxan> luckely usb stacks
[23:25:05 CEST] <buu> haha
[23:25:12 CEST] <onefix> buu: Yeah, but it's to account for the fact that I can be recording up to 3 programs at once ... you wouldn't want to hammer the system with 8 encode processes at once, so I have it run through the script and then delete the "lock file"...which throttles the encoding.
[23:25:12 CEST] <buu> daisy-chained usb ports lighting your house on fire
[23:25:25 CEST] <buu> ah
[23:25:35 CEST] <buu> Recording cable is just so 2009
[23:25:56 CEST] <furq> it would probably be cheaper to just buy two more cheap amd systems
[23:26:00 CEST] <furq> and much faster
[23:26:12 CEST] <muxan> ive got an nice linux setup to record cable
[23:26:19 CEST] <buu> This is clearly less fun furq
[23:26:23 CEST] <onefix> buu: It sometimes means that programs take a while to transcode, but they are available in the "raw" mpeg2-ts stream until they are encoded to MKV files
[23:26:27 CEST] <furq> oh yeah sorry i'm being no fun
[23:26:39 CEST] <onefix> muxan: Using a HDHomeRun?
[23:26:48 CEST] <muxan> most cable streams are mp4 prolly x264 anyway a simple copy will do
[23:27:10 CEST] <buu> or you could just torrent the already encoded files -_-
[23:27:22 CEST] <furq> are you recording three simultaneous programs constantly
[23:27:22 CEST] <muxan> no settop box i mean
[23:27:25 CEST] <onefix> buu: There's no fun in that :)
[23:27:27 CEST] <furq> because it seems like you could just queue them up
[23:27:37 CEST] <furq> +otherwise
[23:27:53 CEST] <onefix> muxan: I'm pretty sure that cable is all still MPEG2-TS since that's what the HDHR outputs
[23:28:34 CEST] <furq> also if this is progressive SD then you should be able to transcode them much faster than realtime
[23:28:36 CEST] <muxan> well same copy applies
[23:29:42 CEST] <muxan> a settop box like a dreambox or something records cable just fine without reencodes
[23:30:11 CEST] <muxan> is prolly cheaper than rpi cluster
[23:30:18 CEST] <TD-Linux> the raspberry pi actually has a hardware h264 encoder too. but it's trash
[23:30:59 CEST] <onefix> furq: Well, that's what I 'm doing now...but presumably putting a CHIP or a rPi or another desktop on the encoding process should improve performance ... I'm not thinking that it will be as fast as a multi core x86 processor, but it should do something ... I also wonder what the cost vs performance difference would be between say a $500 6-core AMD box and 50 CHIPs
[23:31:36 CEST] <furq> well if you already have a 6-core AMD box then that's cheaper
[23:31:46 CEST] <TD-Linux> onefix, I've already done the math, basically zillions of cores wins
[23:31:53 CEST] <TD-Linux> (x86 cores)
[23:32:07 CEST] <TD-Linux> and generally intel beats AMD
[23:32:19 CEST] <muxan> cheapes chips are rpi n they go $30 here
[23:32:59 CEST] <TD-Linux> why am I even giving this advice
[23:33:29 CEST] <DHE> anyone tried encoding with an Intel Phi?
[23:33:31 CEST] <onefix> muxan: The CHIP is $9 and it comes with a 1GHz R8, 4GB  Flash, 512GB RAM, WiFi, and Bluetooth
[23:34:29 CEST] <muxan> i think 3 celerons beats a 100rpis
[23:34:58 CEST] <muxan> n those chips of 1ghz too arm
[23:35:50 CEST] <onefix> muxan: Yeah, the R8 is an ARM processor ...
[23:35:59 CEST] <muxan> indeed
[23:36:24 CEST] <muxan> a cheap celeron is prolly 150
[23:36:46 CEST] <muxan> like those chromebox you can mode to run linux
[23:37:16 CEST] <muxan> modd*
[23:37:43 CEST] <muxan> asus cn62 or so
[23:38:14 CEST] <onefix> I'm going to try transcoding a 30 second MPEG2-TS video on a rPi, a CHIP, my 6-core AMD, and a 4-core Intel and comparing the speeds
[23:39:07 CEST] <muxan> ill be interested in results
[23:39:35 CEST] <muxan> who knows you may proof me wrong
[23:39:59 CEST] <furq> http://sprunge.us/GYQY
[23:40:11 CEST] <furq> that's what i get on an ntsc dvdr on an rpi2 at 1100mhz
[23:42:04 CEST] <furq> compared to ~160fps on an i7 920
[23:44:13 CEST] <muxan> you have a 920 too?
[23:44:21 CEST] <furq> i do
[23:44:28 CEST] <muxan> what are the odds
[23:45:07 CEST] <muxan> i guess bigger than with AMD
[23:45:24 CEST] <onefix> Besides, I think the CHIP is really what the rPi should have been
[23:45:56 CEST] <furq> the rpi would be a much nicer proposition if it had onboard emmc
[23:46:05 CEST] <onefix> furq: https://www.kickstarter.com/projects/1598272670/chip-the-worlds-first-9-computer/description
[23:46:12 CEST] <furq> the thing that really kills it is the sd card
[23:46:21 CEST] <furq> 1gbit ethernet wouldn't go amiss either
[23:47:01 CEST] <muxan> i got a cb2 for $10 more than rpi
[23:47:26 CEST] <onefix> furq: The CHIP doesn't have an ethernet port builtin, it does USB OTG, WiFi, or Bluetooth out of the box
[23:47:48 CEST] <muxan> but everyone jumped on the rpi bandwagon
[23:48:50 CEST] <furq> i don't remember cubieboard's website being this bad
[23:48:58 CEST] <onefix> muxan: Well, the CHIP is also newer ... the R8 is a 1GHz processor, vs the 700MHz processor on the rPi and it has 512M of RAM by default
[23:50:34 CEST] <furq> the rpi2 is 900mhz and clocks easily to 1.1ghz
[23:50:41 CEST] <furq> and it has 1gb ram
[23:51:46 CEST] <onefix> furq: Yeah, but you still need to add an SD card to the rPi :)
[23:51:55 CEST] <muxan> rpi are overrated
[23:52:32 CEST] <furq> also i'm in the uk so the rpi is comparatively cheaper here
[23:53:17 CEST] <onefix> furq: Well, not cheaper than $9 ...
[23:53:24 CEST] <furq> that's why i said comparatively
[23:54:04 CEST] <furq> i can't imagine the onboard storage on the CHIP is any good for that price
[23:54:11 CEST] <furq> but it could hardly be worse than an sd card
[23:54:27 CEST] <onefix> furq: It's NAND
[23:54:38 CEST] <furq> i bought a pretty good toshiba uhs-i sd card and it's still terrible
[23:56:08 CEST] <muxan> terrible compared to sammy 850 pro?
[23:56:23 CEST] <furq> i think all sd cards are terrible for this job
[23:56:50 CEST] <muxan> i guess you right
[23:56:53 CEST] <furq> i wouldn't mind a SATA port on this thing either
[23:57:04 CEST] <muxan> real time stuff is not for sd
[23:57:07 CEST] <furq> one of these days i'll be bothered to get it running over iSCSI
[23:57:58 CEST] <furq> neat, the chip is armv7
[23:58:15 CEST] <furq> that's a big selling point
[23:58:18 CEST] <muxan> still anyone know how i can verify a flac encode?
[23:58:38 CEST] <furq> do you mean a cd rip
[23:58:46 CEST] <furq> you can use accuraterip but i don't know of a way to do it with ffmpeg
[23:59:08 CEST] <muxan> yeah ive also got an old wavpack i want to mux
[23:59:23 CEST] <muxan> but i need to be sure its errorless
[23:59:30 CEST] <furq> cuetools is great for this but it's windows only
[23:59:43 CEST] <furq> i think it runs in wine
[23:59:58 CEST] <furq> http://www.cuetools.net/wiki/Main_Page
[00:00:00 CEST] --- Fri Oct 23 2015


More information about the Ffmpeg-devel-irc mailing list