I am trying to save the Decklink input as a video file (applying specified
below filters) and I am wondering why the extra filter is inserted.
You can find the logs enclosed.
I would be really grateful if you could explain why the filter is auto
inserted there and if there is any option to avoid it.
Do you have any idea how to omit this scaling and get frames uploaded to
cuda device? It seems that hw_upload_cuda needs yuv444p but maybe there
exists another way to do it without the initial scaling performed on CPU
and do the scaling before next filters already on GPU.
ffmpeg-user mailing list
[hidden email] https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email
[hidden email] with subject "unsubscribe".
Regarding changing Decklink output -> it is not configurable.
scale_npp=format=yuv444p is here because I tried to explicitly force the
first scaling to happen on cuda -> did not know that this format is
required by hw_upload, I thought it might be the yadif requirement. But it
did not help and still the first pixfmt change happens on cpu.
deinterlacing -> my fault, by mistake I forwarded already deinterlaced
stream to decklink in the example from the logs, normally I get interlaced
input, that's why deinterlacing is desired