Recherche avancée

Sur d’autres sites (11876)

• ffmpeg with cuda and concat videos [closed]

  15 septembre 2024, par Petr Šimůnek

  I have a ffmpeg command with nvidia hw support that I want to join two videos :

  


  /home/videotest/ffmpeg-dev/ffmpeg -benchmark -hide_banner -loglevel warning -y -hwaccel_device 0 -hwaccel cuda -hwaccel_output_format cuda -i /mnt/video-storage/test/cuda/surfing/intermediate1.mp4 -i /mnt/video-storage/test/cuda/surfing/intermediate2. mp4 -filter_complex 'hwupload_cuda,[0:v:0][0:a:0][1:v:0][1:a:0]concat=n=2:v=1:a=1 [outv] [outa]' -map [outv] -map [outa] -pix_fmt yuv420p -c:v h264_nvenc -preset p6 -profile: v high -force_key_frames 'expr:gte(t,n_forced*5)' -bf 0 -movflags +faststart -b:v 3000k -maxrate 4000k /mnt/video-storage/test/cuda/surfing/game.mp4

  


  I get an error message after starting :

  


  Impossible to convert between the formats supported by the filter 'graph 0 input from stream 0:0' and the filter 'auto_scale_0' [fc#0 @ 0x61ae2b37a1c0] Error reinitializing filters!

  


  I can't figure out how to remove it. Can you help me ? Thank you.

  


  If I use nvdec/nvenc instead of cuda, this error does not appear and the video is created fine.

  


• Use ffmpeg to compile RGBA image sequence into two separate videos (RGB + Alpha)

  27 novembre 2018, par MirceaKitsune

  I plan on using Blender to render animated sequences for a game. I need the videos to have transparent backgrounds in the format expected by its engine. This format involves the transparency being defined as a separate grayscale video of equal FPS and duration. Since splitting the RGB and Alpha channels in Blender is more complicated, I’d prefer doing this directly from ffmpeg.

  The input directory contains an image sequence of png files in RGBA format (eg : 0000.png to 0100.png). What I need ffmpeg to do is compile them into two separate videos : One containing only the RGB channels (leave transparency black) and another containing only the grayscale alpha channel. In the end I’d have something like my_video.mp4 + my_video_mask.mp4.

  I’m familiar with compiling non-transparent image sequences into video using :

  ffmpeg -f image2 -i /path/to/sequence/%04d.png output.mp4

  But I don’t know how to extract the alpha channel and make it a separate video file. What is the simplest ffmpeg command to achieve this result ?

• How can I render frames decoded by FFmpeg using hardware decoding with D3D11 ?

  14 juin 2024, par mercuric taylor

  I have completed the process of decoding a video frame using FFmpeg. The format of the decoded frame is AV_PIX_FMT_NV12. Now, I want to render this frame to the screen using D3D11. My questions are :

  


  

  1. What is the equivalent concept in D3D11 for a decoded frame ? Is it a texture ?

  


  2. I have seen many solutions that convert NV12 data to RGB, but it seems that DX11 does not require this conversion anymore.

  


  3. I just want to display this frame, and since my frame is on the GPU, is there a more convenient way to render directly on the GPU without copying ?

  


  


  Please forgive my not-so-good English. Can anyone provide a reference example ?

  


  I have already referenced this open-source project. texthttps://github.com/balapradeepswork/D3D11NV12Rendering/tree/master/D3D11NV12Rendering
But I don't understand it very well.
Since I don't use dx11 to make game, just for show video, I am looking forward for a more easy solution. It has worried me for weeks, can anyone give me some advice(even good tutorial, some I can found is too old). Thanks sincerely !