On 1/18/2021 9:57 AM, Bartosz Ziemski wrote:
Do you have any idea how to omit this scaling and get frames uploaded to
cuda device?
hwupload_cuda,
scale_npp=format=yuv444p, (why is that there?)
yadif_cuda,
scale_npp=w=1280:h=720:interp_algo=lanczos:format=yuv420p
I'm certainly not well-versed in cuda (and I'm sure someone will correct
me...) but...
That's multiple pixfmt changes, which may not be necessary.
The input is progressive* why run it through a deinterlacer?
*(Stream #0:1, 2, 1/1000000: Video: rawvideo, uyvy422(progressive), 1920x1080)
It just seems to me that the entire filter graph doesn't make sense. What's
the goal?
Later,
z!
_______________________________________________
ffmpeg-user mailing list
[email protected]
https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email
[email protected] with subject "unsubscribe".
