Recherche avancée

Sur d’autres sites (7488)

• Programming filter_complex graph in ffmpeg to enable 'overaly_cuda' with transparency in input frames [closed]

  16 mai 2023, par time4tea

  I'm trying to overlay a full frame image on top of an MP4, in ffmpeg, using as much hardware-assistance as possible.

  


  This is for https://github.com/time4tea/gopro-dashboard-overlay

  


  I have the non-GPU overlay working totally fine, and would like to take advantage of the GPU now.

  


  The inputs to ffmpeg are :

  


  

  • a stream from an MP4 file - with 1 audio and 1 video stream

  


  • frames from the dashboard software, which are supplied as raw RGBA frames, at the same frame size as the movie.

  


  


  I've managed to get the overlay sort of working using the nvidia extensions, but it doesn't work as it uses nv12 frame formats which don't have any idea of transparency (AFAIK) so the overlaid frame completely obscures the movie frame - makes sense.

  


  The incantation that does this is (put on separate lines to try to keep the different sections - inputs / transforms / outputs - obvious) :

  


  ffmpeg -y -report -hide_banner -loglevel info 
-hwaccel cuda -hwaccel_output_format cuda 
-i movie.mp4 
-f rawvideo -framerate 10.0 -s 1024x576 -pix_fmt rgba -i - 
-filter_complex [0:v]null[mp4_stream];[1:v]format=nv12,hwupload_cuda[overlay_stream];[mp4_stream][overlay_stream]overlay_cuda,hwdownload,format=nv12 
-vcodec h264_nvenc 
-rc:v cbr -b:v 20M -bf:v 3 -profile:v high -spatial-aq true 
-movflags faststart 
output.mp4


  


  When i try to use yuva420p (which I believe is the correct format, but happy to be corrected) - in all various different places, for example :

  


  ffmpeg -y -report -hide_banner -loglevel info 
-hwaccel cuda -hwaccel_output_format yuva420p 
-i movie.mp4 
-f rawvideo -framerate 10.0 -s 1024x576 -pix_fmt rgba -i - 
-filter_complex [0:v]null[mp4_stream];[1:v]format=yuva420p,hwupload_cuda[overlay_stream];[mp4_stream][overlay_stream]overlay_cuda,hwdownload,format=yuva420p -vcodec h264_nvenc 
-rc:v cbr -b:v 20M -bf:v 3 -profile:v high -spatial-aq true -movflags faststart output.mp4


  


  I get flavours of the same error message :

  


  Impossible to convert between the formats supported by the filter 'Parsed_null_0' and the filter 'auto_scaler_1'


  


  AIUI - auto scalers are inserted into the pipeline when things don't match resolutions or formats - but here everything should match ?

  


  Using a filter_complex of [1:v]format=yuva420p,hwupload_cuda[overlay_stream];[0:v][overlay_stream]overlay_cuda,hwdownload,format=yuva420p i.e not using a null filter, which I put in just to make quickly changing filters easy, has the same error.

  


  Impossible to convert between the formats supported by the filter 'graph 0 input from stream 0:0' and the filter 'auto_scaler_1'


  


• Image sequence to video - file naming problem ffmpeg

  10 janvier 2024, par user1405884

  im trying to make video from my gopro sequence
but i need to quick rename by shift > select > f2
so the name are a (1), a (2), a (3) instead of GOPR0001, GOPR0002, GOPR0003

  


  

  


  How can i make this new set of renamed files in ffmpeg ?

  


  I believe -i GOPR%04d.jpg

  


  ffmpeg -r 10 -f image2  -start_number 0001 -i GOPR%04d.jpg -vframes 1000 -vcodec libx264 -crf 25  -pix_fmt yuv420p -vf "scale=1366:768:force_original_aspect_ratio=1, pad=1366:768:-1:-1, setsar=1" input.mp4


  


  i think this script need to be changed

  


  ffmpeg -r 10 -f image2  -start_number 2302 **-i GOPR%04d.jpg** -vframes 1000 -vcodec libx264 -crf 25  -pix_fmt yuv420p -vf "scale=1366:768:force_original_aspect_ratio=1, pad=1366:768:-1:-1, setsar=1" input.mp4


  


• Storing high quality video stream (from IP Action Camera) continuously to a Storage Device and issues around Write-Speed bottlenecks

  21 mai 2017, par Aldo

  I am looking to get an Action Camera (Eken H8R, or a GoPro) for a project. Let me explain the scenario I have :

  I will obtain a live stream from the camera on to a Raspberry Pi (over WiFi) as shown here. Next, I want to continuously segment this live stream into 10-15 minute video clips and store them in an external Hard Drive (which is connected to the RPi). I am considering this thread, and the avconvcommand mentioned in the answer by Alexander.

  Now, my concern is write-speed limitations. The video stream would probably have a high bitrate, and might be 4k30fps, in which case file sizes would be huge. Would write speed to disk be a bottleneck ? If so, how will the Pi maintain a buffer to achieve this ? Could I run out of space as the memory stick on the Pi would only be around 8 gigs ? If so, what’s a better alternative ? Please correct me if I’m wrong on these as I’m only a beginner.