For those still technically curious about this performance difference, I would like to point out that ffplay is sending the YUV 4:2:0 decoded video directly to the "display API" at 1.5 bytes per pixel. Kdenlive is using OpenGL for video output, which requires conversion to RGBA, which is 4 bytes per pixel. I was not able to integrate direct YUV output as a view embedded in an app, so I used OpenGL because there is QGLWidget in Qt. Not only is there much more data throughput at 4 bytes per pixel (esp 60 fps!), but there is the image conversion from decoded YUV plus the other MLT overheads mentioned previously.