Hi, I spent a lot of time to try to configure the rendering engine. As you know FFMPEG made a collaboration with the NVIDIA engeneers. From this work is born the new ffmpeg code. Now FFMPEG implements the CUDA engine. I tryed by command line and I notice that the ffmpeg rendering is per SECONDS. The melt rendering is PER HOURS. The quality is very good. Please there is the way to configure the kdenlive interface to operate with FFMPEG rendering engine (that supports CUDA by NVIDIA)? Thanks.

I do not know a way to reroute rendering around MLT today in kdenlive.

As a feature proposal: It would be very nice to have a feature similar to pipe rendering in cinelerra-gg. There you choose a quickly renderable raw format (e.g. yuv or rgb) and as a output filename you choose a named pipe you created before. Then, on command line, you use ffmpeg and read the rawvideo from the named pipe. As you have full power over all ffmpeg commandline switches, you can use hardware encoding on GPU.

Yesterday I got that GPU stuff running and received an overall speed of 220fps (ffmpeg speed measuring) in opposite to 20-30 fps using MLT (720p -> 720p rendering). I have a 5 year old i7-3632qm and I used the integrated intel HD 4000 GPU via VAAPI (which is also implemented by ffmpeg).

Yes! The first thing that one must look for is the power. Now I cannot work with 70 fps. It is not possible. The coders should do something, but in the freeware software this is difficult. Many thanks!

I have seen that kdenlive has got some support on hw acceleraton in the 17.04 release. Check this bug report and solution proposals: https://bugs.kde.org/378832

I did not understand. I gave an eye to cindelerra. I will try to learn it. Kdenlive is simplier but expansive of power. I want to use the CUDA device of course. Many thanks.