[01:10] <synth_> is there a way to add a watermark to a file i'm generating? if so, how would I alter my current command http://pastebin.com/Bs8vUykw ? [01:59] <klaxa> synth_: see the overlay filter and/or google for: ffmpeg watermark [01:59] <klaxa> there are lots of guides [02:18] <synth_> yeah i tried a couple but i couldn't figure out how to do a watermark + text [02:22] <Moussekateer> hi guys. Can someone help me with this issue http://pastie.org/pastes/9009983/text?key=cn8uvmwuelt3laypinhpyq [02:23] <Moussekateer> I see lots of google results for it but all the solutions seem to be of the form 'keep trying different encoding bitrates' [02:25] <Moussekateer> the scripts I'm using have ffmpeg as a dependancy, so it would be nice to find a solution that doesn't require me to rewrite everything [02:31] <c_14> Can you try adding a -loglevel debug? [02:32] <synth_> is it possible to run multiple -vf commands? [02:33] <c_14> Just separate them with commas, or use a -filter_complex [02:33] <c_14> Or read the guide: https://trac.ffmpeg.org/wiki/FilteringGuide [02:46] <Moussekateer> c_14: http://pastie.org/pastes/9011364/text?key=e4zt0p92kfcjpvu5inrlg [02:48] <klaxa> >[mp4 @ 00000000057d87e0] pts (6016) < dts (6704) in stream 0 [02:48] <klaxa> your input is broken [02:49] <synth_> c_14 http://pastebin.com/7z9xVQXf i'm getting a "[AVFilterGraph @ 0x2ed40e0] Too many inputs specified for the "movie" filter." error [02:50] <Moussekateer> klaxa: Broken in what sense? [02:51] <Moussekateer> I'm not sure exactly what that statement means [02:51] <Moussekateer> the pts < dts one [02:51] <klaxa> ah wait a sec... [02:51] <klaxa> the... output is broken? [02:52] <klaxa> can you remux to mkv? just to see if it's the container [02:52] <klaxa> dts means decoding timestamp, pts means presentation timestamp [02:52] <klaxa> in that case the presentation timestamp is smaller than the decoding timestamp [02:52] <klaxa> meaning the frame would have to be displayed before it has been decoded which is impossible [02:54] <c_14> synth_: I'm pretty sure your quoting is off somewhere. [02:54] <Moussekateer> klaxa: Same issue http://pastie.org/pastes/9011775/text?key=uwyazobx1auf5rabrykyq [02:55] <klaxa> can you play back the file without errors? [02:57] <Moussekateer> klaxa: The original? Yes, it plays fine [02:57] <synth_> i'm confused now [02:57] <Moussekateer> seeking and all [02:58] <klaxa> that is weird [03:01] <synth_> http://ffmpeg.org/ffmpeg-filters.html#drawtext-1 for drawtext shows -vf drawtext="params here"; but if i wanted to also do movie params; other sites are showing me to do -vf "movie=params here" (http://www.idude.net/index.php/how-to-watermark-a-video-using-ffmpeg/). is that incorrect? [03:01] <synth_> i'm using their code for placing the watermark on the bottom right [03:02] <synth_> like: -vf drawtext="draw text params here","movie=watermark params here" [03:02] <synth_> since you guys said i could separate the -vf commands with a comma [03:03] <c_14> The problem is that your first "" pair is quoting 1 filter. Your second "" pair is quoting the movie filter _and_ some other stuff. [03:07] <c_14> Try this: http://sprunge.us/SFCe [03:07] <synth_> i got the movie filter from: http://www.idude.net/index.php/how-to-watermark-a-video-using-ffmpeg/ under the "Bottom right corner"; is his example incorrect? i basically just copied that and repathed to my watermark image [03:07] <c_14> I don't see it doing anything with the video I tested it on, but at least it didn't error. [03:08] <synth_> ok one second [03:08] <c_14> synth_: You can either quote each filter individually or the whole filterblock at once, he quoted the whole block, while you quoted the first filter individually _and_ quoted the rest as a block. [03:09] <c_14> You might need to get rid of either the first or the second [in], I'm not sure there are supposed to be two of them. [03:10] <synth_> still getting an error using your example [03:10] <synth_> i tried quoting the whole filterblock for the first as well [03:11] <c_14> Try this one: http://sprunge.us/ibJL [03:11] <c_14> And if that doesn't work, pastebin the result so I can see the error msg. [03:15] <synth_> it worked! [03:15] Action: synth_ flails [03:18] <synth_> thank you so much [03:18] <c_14> No problem. [03:18] <synth_> i owe you a beer [03:33] <synth_> is there a way to show the file frame number versus the frame number starting from 0? (i'm using: Frame: %{n}) [03:34] <synth_> sometimes we start our frame ranges at other numbers like 1001 instead of 0001 [03:40] <klaxa> synth_: see: http://ffmpeg.org/ffmpeg-filters.html#drawtext-1 [03:40] <klaxa> >start_number The starting frame number for the n/frame_num variable. The default value is "0". [03:40] <synth_> ahh okay [03:41] <synth_> thank you! [03:41] <klaxa> incidentally i used the drawtext filter today too :) [03:42] <klaxa> it's easier to draw frame count/pts with drawtext than to write a script to generate a subtitle file [03:48] <Moussekateer> klaxa: Is there any fix I can carry out then, other than resorting to other software? [03:49] <klaxa> none that i know of [03:56] <Moussekateer> mboo [06:41] <orgy> hello? [06:42] <orgy> i want to ask [07:42] <llogan> synth_: you don't need movie filter for overlays [07:49] <ScottSteiner_> klaxa: I tried using that guide to hardcode the subs, but it didnt work [16:08] <quup> How do I resize a movie to fixed height and width by first scaling to fit the height and then cropping sides in one go? [16:08] <quup> without any stretching that is [16:09] <klaxa|work> quup: use the scale and then the crop video filter [16:11] <klaxa|work> i.e.: -vf scale=-1:480,crop=w=640 [16:11] <quup> klaxa|work: so like -vf scale=-1:480,crop=w=480 [16:11] <quup> :D thanks! [16:11] <klaxa|work> i think that should work, i have not tested it [16:11] <quup> I'll give it a try [16:15] <quup> klaxa|work: works great, thanks [16:17] <klaxa|work> if you want to crop centered you will have to do some math though [16:18] <klaxa|work> like (i_w-400)/2 for the left side and i_w-((i_w-400)/2) for the right side [16:18] <klaxa|work> maybe [17:10] <ronlut> Hey guys, I'm building an android application, and I need to use some ffmpeg commands. I've got a source video and I need to edit it: 1) Cut only the last 7 seconds. 2) Add a watermark(logo) on the top left side of the video.(00:00-00:07) 3) Add a video inside the source video on the bottom right side of the video.(00:00-00:07) 4) Add another video to the edited video from steps 1-3 (00:07-00:10) Can I do it in one ffmpeg command? [17:10] <ronlut> The commands I used so far are: http://pastebin.com/nPuuuHh7 But I have a problem with it: I hear the sound of the video in step 3 instead of the main video. Thanks a lot! [17:26] <deni> .close [19:02] <synth_> is it possible when using the movie= -vf filter to add a semitransparent png overlay to a video? it looks like the transparent png i'm using is fully opaque and not antialiased around the edges [19:24] <llogan> synth_: why are you using movie? [19:35] <synth_> c_14 helped me set it up lsat night [19:35] <synth_> llogan [19:36] <synth_> yes i know i was just preparing it for you [19:36] <synth_> everything works fine [19:36] <synth_> llogan http://pastebin.com/kn4D2Y2s [19:37] <synth_> no errors [19:37] <llogan> you forgot the console output [19:42] <synth_> llogan http://pastebin.com/dfUZ6EpF [19:42] <llogan> thanks [19:42] <llogan> can you provide your png file? [19:43] <synth_> the watermark png file? [19:43] <llogan> yes [19:44] <llogan> you have two instances of "-s 1280x720" outside of your filtergraph. what is the purpose of the "-s 1280x720"? [19:44] <synth_> http://mellow.me/tmp/bigblock_watermark.png [19:44] <synth_> i thought the first -s was the input resolution and last -s was the output resolution [19:46] <synth_> example of the output mov file it generated: http://mellow.me/tmp/Club_Fracture_V4_.mov [19:47] <llogan> your main input is already 1280x720 and the output will inherit that same size [19:47] <synth_> i still need to implement the start_number parameter to get the frame number part working [19:47] <synth_> oh okay [19:47] <synth_> i'll remove the second one then [19:47] <llogan> you don't need either of them [19:48] <synth_> so if the input was 1080, and someone wanted to render out a 720, could i implement them? [19:48] <synth_> or would the source frames need to be 720 [19:49] <llogan> you can scale any input. i'll make an example once i finish something [19:49] <synth_> ok [19:49] <synth_> thanks man [19:50] <llogan> also, you can use the glob pattern instead of -start_number. might be easier for you [19:50] <llogan> example: http://ffmpeg.org/ffmpeg-formats.html#Examples-3 [19:51] <synth_> okay i didn't know about that [19:51] <synth_> thanks [19:55] <llogan> -vendor is ignored because it's a provate option for the prores_ks encoder only, AFAIK [19:56] <llogan> *private [20:04] <llogan> synth_: http://pastebin.com/XpPVjkT5 [20:04] <llogan> i'm not sure if you needed that probesize [20:04] <llogan> the scale in the first filterchain shows that you can scale any input at any time in the filtergraph [20:05] <llogan> the overlay filter needed its format option because by default it uses yuv420: http://ffmpeg.org/ffmpeg-filters.html#overlay [20:06] <llogan> (it is not the same thing as the format filter used at the end) [20:07] <llogan> but you won't need "scale=1280:-1,". it's just there as an example [20:25] <synth_> i was reading somewhere that probesize is good for frame by frame scrubbing using arrow keys [20:27] <llogan> it's used to "probe" the input to get stream information [20:28] <llogan> what you're probably referring to is "-g", but i don't think it does anything with mjpeg [20:30] <synth_> i'm not sure, it's on https://trac.ffmpeg.org/wiki/vfxEncodingGuide [20:30] <synth_> for vfx related ffmpeg renders [20:31] <llogan> that's using an encoder that uses "group of pictures", but mjpeg is just a series of images, and therefore does not need, nor can not use -g [20:32] <llogan> i suppose i should read that verbose wiki page someday [20:32] <llogan> so many words [20:32] <synth_> yeah i've just been skimming over it [20:35] <llogan> did my example fix the shitty looking png overlay? [20:37] <synth_> let me see if i can restructure my code without breaking anything... [20:37] <synth_> the pattern glob portion isn't going to be easy to fix [20:37] <synth_> so i'll have to leave that part out [20:38] <synth_> we have some renders that have very identical names but are different passes all dumped into one directory [20:38] <synth_> s/very/almost [20:40] <llogan> you can still use glob: -i "foo*.png" [20:57] <brontosaurusrex> hi ~ [20:58] <synth_> llogan i think your example has an issue with quotes [21:00] <synth_> llogan this is what i'm getting: http://pastebin.com/PzDfhK9j [21:00] <synth_> i refactored your code to fit into what i already had and it looks like encapsulating the quotes within quotes for the -filter_complex for the draw_text parameter messed things up [21:11] <llogan> synth_: s/drawtext="fontfile/drawtext=fontfile and s/10"[bg]/10[bg] [21:16] <synth_> llogan new error: http://pastebin.com/sG7D14au [21:22] <llogan> synth_: i forgot to add -map "[out]" in the example [21:22] <llogan> as an output option [21:22] <synth_> oh okay, does that come after the -filter_complex ? [21:25] <synth_> wait nevermind [21:25] <llogan> you can probably place it anywhere as long as it is an output option, but for humans it would make more sense to place it after the filtering [21:26] <synth_> ok so just rendered out another test and the png doesn't look any better, same as before. [21:27] <llogan> you could remove [out] from the filtergraph and then omit -map "[out]", but i prefer to explicitly llabel things so i know exactly what is going on instead of relying on defaults. [21:27] <synth_> wait a second [21:27] <synth_> it looks better as long as black isn't behind the png [21:27] <synth_> let me upload an example... [21:30] <synth_> llogan : http://mellow.me/tmp/3000_0010_lgt_scenestate1_v0001.mov so the logo now is jagged over black, but as soon as color starts coming in, it looks as expected [21:31] <llogan> weird [21:31] <llogan> i can't duplicate the issue, but i don't have similar inputs [21:32] <synth_> i could upload the frames, and watermark if you'd like? [21:32] <llogan> what is the file size for each typical frame? [21:33] <synth_> can range from a 2-10mb to 200-300mb per frame [21:34] <synth_> these frames max around 9-10mb [21:35] <llogan> i guess the first ~48 will suffice [21:35] <llogan> maybe even just ~38-48 [21:36] <synth_> ah crap sorry just uploaded all of the frames lol [21:37] <synth_> ~340mb [21:37] <synth_> want me to redo it? [21:37] <synth_> er ~370mb [21:38] <llogan> no, i'll download it while i'm taking my break [21:40] <synth_> ok http://mellow.me/tmp/9999_testproject_3000_0010_lgt_cam003_v0001__frames.zip [21:40] <synth_> do you work for ffmpeg? [21:43] <llogan> FFmpeg is made up of volunteers. i work for a small media business where i use ffmpeg very often. [21:44] <llogan> i'll be back in about an hour [21:48] <synth_> okay [22:40] <ronlut> Hey guys, I'm building an android application, and I need to use some ffmpeg commands. I've got a source video and I need to edit it: 1) Cut only the last 7 seconds. 2) Add a watermark(logo) on the top left side of the video.(00:00-00:07) 3) Add a video inside the source video on the bottom right side of the video.(00:00-00:07) 4) Add another video to the edited video from steps 1-3 (00:07-00:10) Can I do it in one ffmpeg command? [22:40] <ronlut> The commands I used so far are: http://pastebin.com/nPuuuHh7 But I have a problem with it: I hear the sound of the video in step 3 instead of the main video. Thanks a lot! [22:56] <groundnuty> hey, I'm having trouble with ffmpeg ffmpeg -pattern_type glob -i '*.png' $MOVIE [22:56] <groundnuty> it takes files like ls would print them [22:56] <groundnuty> and I would pref it would take them as ls -v would print thme [22:56] <groundnuty> I would not like to reneme them all [22:57] <groundnuty> i a concat option an answer here? [22:57] <quup> groundnuty: what's the difference between ls and ls -v in this case? [22:57] <quup> oh numerical sort [22:59] <quup> groundnuty: if you replace "-pattern_type glob -i '*.png'" with just -i '%d.png' does that give correct order? [22:59] <groundnuty> quup: no, the numbers are crazy [23:00] <groundnuty> quup: tried play with %d [23:00] <groundnuty> but they are like prefix.timestamp.soemthing [23:00] <groundnuty> prefix.timestamp.soemthing.png ;p [23:00] <groundnuty> I thought that glob worked till prfix was in range of 0-9 [23:01] <groundnuty> when 10 apeared it was taken after 1 [23:01] <quup> groundnuty: maybe there is an option to read input filenames from a file [23:01] <quup> so you could just ls -v > input_files.txt and then ffmpeg --magic-thingy=input_files.txt [23:01] <sacarasc> groundnuty learns the importance of leading zeros. [23:02] <groundnuty> sacarasc: I wont comment on those filenames - not my program generates them :( [23:03] <groundnuty> I dont see the option that would read input files from external file [23:03] <groundnuty> sacarasc, quup am I missign something? [23:04] <quup> groundnuty: don't think so, can't find any such option either [23:05] <quup> I guess I shouldn't be too surprised ffmpeg doesn't have a -playlist option :D [23:06] <quup> groundnuty: oh! it exists, -f concat -i thefilenamesinthisone.txt [23:06] <quup> groundnuty: http://trac.ffmpeg.org/wiki/How%20to%20concatenate%20%28join,%20merge%29%20m... [23:06] <groundnuty> quup: I'm reading about it [23:06] <groundnuty> I'm just not sure because its all about wav [23:06] <groundnuty> in my case I have png and output mp4 [23:07] <quup> it's only getting the input files in order that matters [23:07] <quup> rest is just throwing them at the right filters [23:08] <quup> or maybe I missunderstand concat... [23:08] <quup> oh well [23:32] <llogan> synth_: try changing format in overlay from rgb to yuv422. also that maked the last format filter redundant. [23:33] <llogan> *makes [23:34] <llogan> ronlut: without the complete console output helping you will be challenging [23:41] <synth_> llogan: cool! that worked! [23:41] <synth_> thank you! [00:00] --- Thu Apr 10 2014