[Ffmpeg-devel-irc] ffmpeg.log.20150906

[burek]

[01:40:34 CEST] <tommd> On recent macbooks (early 2015, 13' pro) using AVFoundation for input I and at least one other person get this horrible error.
[01:42:02 CEST] <tommd> "Selected framerate (29.970030) is not supported by the device."
[01:43:15 CEST] <tommd> Commenting out the call to 'av_q2d' and instead setting the framerate to a hardcoded correct value of 30 makes it work (For me, obviously)
[01:43:45 CEST] <tommd> Where does the 'framerate' field get set?
[01:44:22 CEST] <c_14> -framerate input option
[01:53:54 CEST] <tommd> c_14: I'm using ffmpeg as a library, not a command line tool.
[01:54:34 CEST] <c_14> The framerate AVOption
[01:54:55 CEST] <tommd> So the context comes from the next_device() call, which walks the linked list created by *_register_all()
[01:55:10 CEST] <tommd> Still not clear what created the AVRational that generates the 29.97 value.
[01:56:07 CEST] <c_14> The AVFContext has a framerate member
[01:56:09 CEST] <c_14> set that
[02:07:37 CEST] <tommd> c_14: Shouldn't it be set when I select the input device?  Is this not a bug?
[02:07:51 CEST] <tommd> (an FFMPEG library bug, vs my program's bug)
[02:30:29 CEST] <c_14> I think you have to av_opt_set_q(ctx, "framerate", AVRational val, AV_OPT_SEARCH_CHILDREN); or something
[02:30:43 CEST] <c_14> Not entirely sure though
[03:36:07 CEST] <Nanashi> What does -g mean?
[03:37:07 CEST] <Nanashi> It's not explained in https://ffmpeg.org/ffmpeg.html or -help
[03:37:26 CEST] <c_14> It's in -formats
[03:37:28 CEST] <c_14> It's the gop size
[03:37:38 CEST] <Nanashi> k
[03:37:48 CEST] <c_14> Eh, -codecs that is
[07:21:46 CEST] <salviadud> I want to stretch a video
[07:22:15 CEST] <salviadud> from 528x480 to 848x480
[07:22:32 CEST] <salviadud> I've used several commands that don't seem to work
[07:22:36 CEST] <salviadud> with -vf scale
[07:22:55 CEST] <salviadud> I'm thinking I might need to output all frames to independent png files
[07:23:16 CEST] <salviadud> and then mod them with a script, but I still can't get ffmpeg to rescale my image
[07:29:42 CEST] <Nanashi> Is  -vf scale=848:480  not doing anything?
[07:44:04 CEST] <salviadud> I'm trying out this one right now
[07:44:39 CEST] <salviadud> ffmpeg -i input.avi -s 848x480 -c:a copy output.avi
[07:44:48 CEST] <salviadud> I'm waiting for it to finish.
[07:45:11 CEST] <salviadud> but, maybe that'll just add some black bars to the sides
[07:45:13 CEST] <salviadud> I hope not
[07:45:38 CEST] <salviadud> Nanashi, I'm sorry I took so long, I had to get some chow
[08:26:16 CEST] <salviadud> I just want to know if ffmpeg can do it
[08:26:27 CEST] <salviadud> That last command you gave me, didn't work either
[08:26:44 CEST] <salviadud> I want to stretch the video completely, the proportions are all wrong
[10:07:29 CEST] <durandal_170> salviadud: have you read documentation?
[10:14:35 CEST] <Nanashi> One a note, 16:9 is closer to 854 instead of 848, so if by "black bars" he means there's a few pxiels of black on top and bottom when fullscreened, then there's the problem.
[12:17:25 CEST] <salviadud> I have a nice setup going
[12:17:32 CEST] <salviadud> Except for that gimp plugin
[12:17:38 CEST] <salviadud> it takes forever to do its job
[12:18:08 CEST] <salviadud> I can't seem to stretch the video content with ffmpeg
[12:18:14 CEST] <salviadud> I'm not sure where to look
[12:18:40 CEST] <salviadud> top shows only one processor core with gimp. so, its a slow workaround
[12:18:46 CEST] <salviadud> other than that, ffmpeg is awesome
[15:38:56 CEST] <debunez> does anyone know what libraries from FFMpeg would be needed to just grab the frames from a video to be rendered using lets say OpenGL or DirectX? Was just wondering if i need to link with ALL of them...
[15:41:49 CEST] <BtbN> Depends on what you want to do
[15:42:26 CEST] <BtbN> avcodec decodes video, avformat demuxes containers, swscale converts color formats and scales
[15:42:33 CEST] <BtbN> So you most likely want to use at least those
[15:43:25 CEST] <debunez> okay sounds good.
[16:48:42 CEST] <DHE> is -sws_flags a global setting? I need to know how it interacts with -filter_complex which contains `[input]scale=...[output]` clauses
[16:54:04 CEST] <tacpilot> working with a linux cluster but each node has very limited resources.. trying to leverage parallel processing. sometimes need to split a single input into many outputs. if I decode the input and pipe it out to the gigabit network, is there any way to sync the remote nodes ? UDP is outta the question I must retain all the data so need some way to sync transcoding across the network..   any suggestions ???
[16:56:23 CEST] <tacpilot> i guess UDP would be ok if there was some other mechanism in place to guarrentee sync
[17:21:07 CEST] <DHE> what do you mean by sync? you producing real-time streams and need them to be all time-synced?
[17:24:37 CEST] <tacpilot> since the nodes have limited resources .. i fear buffer overrun .. i need one machine to decode the footage and send a stream to the network .. i need multiple nodes to take their designated time slices and encode their segments.. is a production studio so can not loose any frames
[17:27:11 CEST] <tacpilot> as is setup .. one machine grinds for 10 hours while the other 15 do nothing ... looking for a way to leverage parallel processing without data loss
[17:30:14 CEST] <tacpilot> hoping there is some mechanism that can keep them in step .. so if buffers get full .. the decoder box will pause to let the encoder boxes catch up ... then continue .. is the general concept i am picturing in my head but not sure what is avail ?
[17:34:28 CEST] <DHE> I see now... never actually tried that myself. I'd sooner suggest encoding 16 different videos at the same time instead
[17:40:51 CEST] <tacpilot> wish that was an option .. this is a stateless cluster so most nodes have No harddrives ..most of the machines have less than 2 gigs of ram and single core cpu's around 1.5 ghz ... there are only a couple nodes with more than 4 gigs or ram and multi core cpu's .. this is a render farm where the lessor nodes are already designated tasks suitable for they specs, but when u send 20 clips to the farm and each clip takes 5-10 hours for th
[20:55:08 CEST] <mguerret> does anyone know a good way to get the duration for each frame? I know i can get the full duration of the video from AVFormatContext, but what about per frame duration?
[20:55:46 CEST] <mguerret> should i just take that full video duration and divide it by number of frames? And then another question, is that duration in milliseconds?
[20:57:07 CEST] <c_14> (next frame pts - current frame pts) / time_base
[21:00:53 CEST] <mguerret> k, i will try that. thanks
[21:30:01 CEST] <mguerret> is there a special way to extract the pts from an AVFrame object? They all seem to be garbage negative values or 0?
[21:41:08 CEST] <c_14> Are you looking at pts or pkt_pts ?
[21:43:53 CEST] <c_14> If the values are garbage though, you have to assume cfr input and generate them yourself.
[22:11:33 CEST] <ChazNunz> Hello
[22:14:00 CEST] <ChazNunz> Is it possible to overaly a "progress bar" on a video.  This progress bar would increment from left to right during playback, but it would be part of the video, not part of the player.
[22:16:45 CEST] <c_14> If you have the bar as a video (not necessarily of equal length to the other video), or as a subtitle stream (something like ass) then you can overlay it onto the video but I don't know of any filter that will let you generate the progress bar itself.
[22:19:23 CEST] <sfan5> drawbox filter with an expression that calculates percentage?
[22:20:12 CEST] <durandal_170> geq, but slow and rgb only
[22:20:25 CEST] <ChazNunz> c_14 and sfan5:  Thanks.  I will investigate.  I am already using lots of drawtext, so i'll look into drawbox first.  C_14 idea could work too.
[22:22:28 CEST] <durandal_170> drawbox is missing pts, etc in expression evaluation
[22:26:06 CEST] <ChazNunz> durandal_170: geq looks like the best choice because drawbox has not pts.  Thanks.  I'll go learn.
[00:00:00 CEST] --- Mon Sep  7 2015