Recherche avancée

Sur d’autres sites (6038)

• FFmpeg what exactly is the filtergraph pipeline like during transcoding ?

  8 septembre 2017, par Jeff Gong

  I have been studying the source code for FFmpeg to attempt to understand its threading model and how it processes inputs. For example, when I run a command like :

  ffmpeg -i video.mp4 -s hd720 -c:v libx264 --preset medium -c:a aac -profile:v main -r 60 -f null /dev/null

  The input itself is irrelevant, but I am trying to understand how the transcoding pipeline works. In the source code, I see that the main steps occur in the functions transcode and transcode_step.

  It seems like for a single input, a single frame is read in, decoded, encoded, and written out. The process is obviously very complex but what I am really not understanding is what FFmpeg is doing when it attempts to build out a filtergraph. For example, in transcode_step of ffmpeg.c, there is the following code that happens right after an output stream has been selected :

  if (ost->filter && !ost->filter->graph->graph) {
  
      if (ifilter_has_all_input_formats(ost->filter->graph)) {
  
          ret = configure_filtergraph(ost->filter->graph);
  
          if (ret < 0) {
  
              av_log(NULL, AV_LOG_ERROR, "Error reinitializing filters!\n");
  
              return ret;
  
          }
  
      }
  
  }

  Does this only apply if I specify a specific series of filtering options to FFmpeg, like the one in this link ? For the sample command I input above, is this code still executed ?

  One last other question I had was for the case where I run an FFmpeg instance with a single input but multiple outputs (perhaps different variants for transcoding). In this scenario, does a single phase of transcode_step take in an input frame and send that frame through decoding and encoding for only a single one of the outputs ? Or does it take a frame at a time and process this frame for each of the outputs we have specified ?

• cpu.c:253 : x264_cpu_detect : Assertion

  12 octobre 2017, par user6341251

  environment :
  ubuntu 16.04_x64 server
  install ffmpeg through apt-get install
  python 3
  

  when I try

  from moviepy.editor import *
clip = VideoFileClip("/root/video.mp4")
clip.ipython_display(width=280)

  Traceback (most recent call last) :
  File "", line 1, in
  File "/usr/local/lib/python2.7/dist-packages/moviepy/video/io/html_tools.py", line 219, in ipython_display
  center=center, rd_kwargs=rd_kwargs, **html_kwargs))
  File "/usr/local/lib/python2.7/dist-packages/moviepy/video/io/html_tools.py", line 97, in html_embed
  clip.write_videofile(**kwargs)
  File "", line 2, in write_videofile
  File "/usr/local/lib/python2.7/dist-packages/moviepy/decorators.py", line 54, in requires_duration
  return f(clip, *a, **k)
  File "", line 2, in write_videofile
  File "/usr/local/lib/python2.7/dist-packages/moviepy/decorators.py", line 137, in use_clip_fps_by_default
  return f(clip, *new_a, **new_kw)
  File "", line 2, in write_videofile
  File "/usr/local/lib/python2.7/dist-packages/moviepy/decorators.py", line 22, in convert_masks_to_RGB
  return f(clip, *a, **k)
  File "/usr/local/lib/python2.7/dist-packages/moviepy/video/VideoClip.py", line 349, in write_videofile
  progress_bar=progress_bar)
  File "/usr/local/lib/python2.7/dist-packages/moviepy/video/io/ffmpeg_writer.py", line 216, in ffmpeg_write_video
  writer.write_frame(frame)
  File "/usr/local/lib/python2.7/dist-packages/moviepy/video/io/ffmpeg_writer.py", line 178, in write_frame
  raise IOError(error)
  IOError : [Errno 32] Broken pipe

  MoviePy error : FFMPEG encountered the following error while writing file temp.mp4 :

  ffmpeg : common/cpu.c:253 : x264_cpu_detect : Assertion ` !(cpu&(0x0000040|0x0000080))’ failed.

  what happend ?

  ---

  @Ronald S. Bultje
  

  I am using a virtual machine

  processor : 0
  vendor_id : GenuineIntel
  cpu family : 6
  model : 13
  model name : QEMU Virtual CPU version (cpu64-rhel6)
  stepping : 3
  microcode : 0x1
  cpu MHz : 3504.000
  cache size : 4096 KB
  physical id : 0
  siblings : 1
  core id : 0
  cpu cores : 1
  apicid : 0
  initial apicid : 0
  fpu : yes
  fpu_exception : yes
  cpuid level : 13
  wp : yes
  flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pse36 clflush mmx fxsr sse sse2 ss syscall nx pdpe1gb rdtscp lm rep_good nopl eagerfpu pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm fsgsbase bmi1 avx2 smep bmi2 xsaveopt
  bugs :
  bogomips : 7008.00
  clflush size : 64
  cache_alignment : 64
  address sizes : 39 bits physical, 48 bits virtual
  power management :
  

• ffmpeg to dvb-c. Need some advices

  12 octobre 2017, par pentarhh

  I’d like to get some advices about ffmpeg transcoding and broadcasting it to DVB-C environment.
  Have a server, based on Linux with external GPUs based on Nvidia P4000. Target of this machine is getting live streams at HLS, then it transcode files to multicast UDP. After transcoding multicast TS gets to receiver (Sumavision EMR 3.0), then modulates into DVB-C with QAM-card for next transmission to fiber.

  So, there is a strange trouble. In IPTV such stream have no visible troubles. Analyzing TS, saw an errors with PCR accuracy (>500 ms ’cause of pseudo-CBR) and very rare discontinuity errors, that could be based on source errors.
  Then I take it to Sumavision and apply inbound parameters (just remultiplexing this stream with making loyal reference, buffer size values and CBR output values). Analyzer sees no problem, after putting in QAM-modulator customize output streams, such as PSI/SI tables, frequency, symbol rate etc.

  After I see the result at some test TVs. First one shows this channel great, with no visible troubles (no artefacts, no asynchronous sound). Second TV has a problem of video, it looks like a lost frames. Sometimes sound interrupts for a milliseconds. But there is no freezes
  ’Cause of ETR 101 290 passes well, may be this model of TV decodes this service incorrect. Example of video record is here.

  Here is the code of ffmpeg. May be, I missed some common parameters, that may get the output is better for such tuners. Can you see it’s alright ?

  sudo -u nobody ffmpeg -threads 0 -v warning -re -hwaccel cuvid -hwaccel_device 2 -c:v h264_cuvid -deint 0 -i http://x.x.x.x/playlist.m3u8 -bsf:v h264_mp4toannexb -map 0:0 -map 0:1 -vcodec h264_nvenc -gpu 2 -cbr true -sc_threshold 0 -filter:v scale_npp=960:540:interp_algo=lanczos -vprofile baseline -b:v 1000k -bufsize 1400k -maxrate 1400k -minrate 900k -force_key_frames 1 -g 50 -bf 0 -refs 1 -r 25 -strict experimental -c:a aac -b:a 64K -af aresample=44100 -ac 2 -flags -global_header -f segment -segment_format mpegts -segment_time 10 /path1/segment-1507744758-%06d.ts -map 0:0 -map 0:1 -vcodec h264_nvenc -gpu 2 -cbr true -sc_threshold 0 -vprofile high -b:v 8000k -bufsize 400k -maxrate 9000k -minrate 7200k -force_key_frames 1 -g 50 -bf 2 -refs 1 -r 25 -strict experimental -c:a aac -b:a 192K -af aresample=44100 -ac 2 -flags -global_header -f tee 
  
  [f=segment:segment_format=mpegts:segment_time=10]/path2/segment-1507744758-%06d.ts|[f=mpegts:muxrate=9333k:mpegts_pmt_start_pid=0xB4A:mpegts_start_pid=0xB4B]'udp://y.y.y.y:1234?buffer_size=0&overrun_nonfatal_option=1&pkt_size=1316&ttl=15'