[02:27:38 CEST] <nicolas17> my theory was right, the video is just remuxed [02:28:12 CEST] <nicolas17> I see stuff like consecutive audio and video frames being swapped, audio dts and pts being slightly different... but frame hashes all match [05:12:45 CEST] <KombuchaKip> Hey everyone. The ffmpeg pastebin subdomain isn't working for me, so I posted my question on migrating from an older libav API to ffmpeg 4 here: https://pastebin.com/Ezp8sNVE [11:21:34 CEST] <stn616> is this ffmpeg support [11:22:17 CEST] <fella> Topic for #ffmpeg: Welcome to the FFmpeg USER support channel [11:28:05 CEST] <stn616> sorry I know this is not arecord but hw:1 is not recording, a search on the interwebs told me to use dump-hw-params but I got a lot of formats, so I have no idea which should I use as -sample-fmt [11:29:56 CEST] <stn616> I'm using a focusrite interface which in the box reads can process 24bit/192khz which is what I want [11:30:44 CEST] <stn616> I only have alsamixer not pulse [11:55:59 CEST] <stn616> ok... so should I choose little endian or big endian? [18:27:14 CEST] <Accord> is it possible to stream in real time from one file to another on my local disk? [18:27:43 CEST] <Accord> I've got this script which reads from a wav file which under normal circumstances is written to in real time [18:27:50 CEST] <Accord> I want to simulate this if possible [18:31:07 CEST] <Accord> nvm, found it [18:42:13 CEST] <vifino> Hey. I made a small program using the libav* C APIs. Converts anything into a format of mine. https://github.com/shinyblink/fhconvs/blob/master/src/2fh.c [18:43:05 CEST] <vifino> Running into bottlenecks though. After setting the codec context to be multithreaded, sws_scale is one of the major bottlenecks. [18:43:55 CEST] <vifino> Any simple way to do frame-based multithreading with sws_scale? [18:44:35 CEST] <vifino> Latency is no problem here, so slice-based threading is not needed. Order is, obviously. [18:49:33 CEST] <furq> nothing in lavfi supports frame threading iirc [18:50:34 CEST] <atomnuker> vifino: you can do your own threading by separating the sws from the decoding loop [18:56:25 CEST] <vifino> alright. so i'd avcodec_receive_frame into a bunch of different frames, call sws_scale on them in a thread pool of sorts? [18:57:10 CEST] <vifino> is the SwContext threadsafe? [19:03:29 CEST] <atomnuker> not sure, but contexts are cheap [20:20:58 CEST] <TheWild> hello [20:21:51 CEST] <TheWild> how to trim zeroes from beginning and end of .wav files (just like picture auto-crop by color)? [20:25:03 CEST] <furq> !filter silenceremove @TheWild [20:25:03 CEST] <nfobot> TheWild: http://ffmpeg.org/ffmpeg-filters.html#silenceremove [20:33:25 CEST] <TheWild> thank you. I haven't tested it yet but after deeper reading it seems to be exactly what I was searching for.hunting precisely for zeroes. [20:34:20 CEST] <TheWild> interesting is that filter accepts named arguments, but at this link someone defined it like silenceremove=1:0:-50dB https://stackoverflow.com/questions/25697596/using-ffmpeg-with-silencedetect... [20:34:41 CEST] <furq> you can use either with any filter [20:35:06 CEST] <furq> or a combination of the two [20:35:07 CEST] <TheWild> ah, those are the *first* three arguments [20:35:11 CEST] <furq> yeah [20:35:24 CEST] <TheWild> thanks [20:36:44 CEST] <TheWild> I regret I didn't know ffmpeg a couple of years ago. For one school presentation I had a badly scaled video (with really thick black border) and not even commercial software could fix that, lol. [21:13:27 CEST] <durandal_1707> TheWild: use named arguments always, as i added more options to that filter in latest master [21:14:06 CEST] <durandal_1707> also for removing silence at and you need also areverse filter twice [21:14:13 CEST] <durandal_1707> s/and/end [22:58:47 CEST] <mixfix41> bent [22:58:54 CEST] <mixfix41> oh wrong window [23:17:56 CEST] <mixfix41> ?oh the ffmpeg documentation for capturing x11 on desktop it uses quite a bit a space but i broguth taht up in here once before i shuold sign up for ffmpeg forum and post o [00:00:00 CEST] --- Mon Oct 22 2018
participants (1)
-
burek