I have been working with below environment.

==>Ubuntu 16.04.3

==>FFmpeg :3.4.2

-- configuration: --prefix=/usr/local/ffmpeg_new/ --enable-cuda --enable-cuvid --enable-nvenc --enable-nonfree --enable-libnpp --extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64 --enable-libvpx --enable-libvorbis --enable-libfreetype

==>Cuda 9.1.85

==> GPU: GeForce GTX 1080ti

My purpose is to accelerate video modifications via GPU source since there is high rate traffic on my media server.

Here is the main steps of my process:

1.Split video and audio of the caller.

time ffmpeg -y -i 230087_caller.webm -vn -ab 256 230087_caller.wav

2.Split video and audio of the callee

time ffmpeg -y -i 230087_callee.webm -vn -ab 256 230087_callee.wav

3.Mux original caller video with callee audio to store the conversation.

time ffmpeg -y -i 230087_caller.webm -i 230087_callee.wav -filter_complex '[0:a]aformat=sample_fmts=fltp:sample_rates=44100:channel_layouts=stereo,volume=0.5[a1]; [1:a]aformat=sample_fmts=fltp:sample_rates=44100:channel_layouts=stereo,volume=0.5[a2]; [a1][a2]amerge,pan=stereo|c0

4.Add time fontfile to the caller_temp fil which fails with below error.

time ffmpeg -y -i caller_temp.webm -vf drawtext=fontfile=/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf:x=8:y=8:fontsize=16:fontcolor=yellow@1:expansion=strftime:basetime=1518172680000000:text='%Y-%m-%d %H-%M-%S' -strict -2 -shortest -c:a libvorbis -c:v h264_nvenc final_font_test.webm

[webm @ 0x29e8540] Only VP8 or VP9 video and Vorbis or Opus audio and WebVTT subtitles are supported for WebM. av_interleaved_write_frame(): Invalid argument Error writing trailer of output_temwp.webm: Invalid argument


Here is the details of my webm file :

Stream #0:0(eng): Video: vp8, yuv420p, 640x480, SAR 1:1 DAR 4:3, 27.58 tbr, 1k tbn, 1k tbc (default) Metadata: title : Video Stream #0:1(eng): Audio: opus, 48000 Hz, stereo, fltp (default)


I have take a look at the nvidia video encode/Decode support matrix,my gpu seems should be supporting this conversion.



However,when I have changed the output file format to mp4/avi/mpeg GPU can be used without any issue.Yet ,webm is important for us as the other container file size becomes risky with storage perspective of view.

I will be appreciated if you share your comments and feeedbacks to proceed with this issue.

Thanks in advance. Regards

  • You're using a H264 encoder which WebM container doesn't support. You need a VP8/9 encoder that make use of nvidia hardware, but I don't see one available. – Gyan Feb 28 '18 at 6:09
  • Thank you Mulvya for your feedbacks . Do you have any recommendation in regards which hardware can support vp8/vp9 encoding instead of nvidia geforce ? Or do you have any recommendation to get the best output performance with this existing GPU ,I mean different type of output container. – oktay eşgül Feb 28 '18 at 7:18
  • You can save to MKV. – Gyan Feb 28 '18 at 7:27
  • Any comment for vp8/vp9 supporting GPU hardware. – oktay eşgül Feb 28 '18 at 7:30
  • On linux, you can use VAAPI to use Intel GPUs for VP8/9. – Gyan Feb 28 '18 at 7:31

According to the Support Matrix NVENC only supports AVCHD (H.264) and HEVC (H.265).

You can use NVDEC to decode VP8/9 given a compatible GPU.

For hardware supported encoding of VP8/VP9 please check VAAPI encoder support in FFMPEG.

To use the default decoder for some input, then upload frames to VAAPI and encode with VP9 and default settings:

ffmpeg -vaapi_device /dev/dri/renderD128 -i input.mp4 -vf 'format=nv12,hwupload' -c:v vp9_vaapi output.webm
  • The Mesa VAAPI driver uses the UVD (Unified Video Decoder) and VCE (Video Coding Engine) hardware found in all recent AMD graphics cards and APUs. – der_michael Jan 8 '19 at 5:26
  • Quicksync supports VP8 encoding since Braswell and VP9 since Apollo Lake, see: trac.ffmpeg.org/wiki/Hardware/QuickSync – der_michael Jan 8 '19 at 5:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.