Jump in Smart rendered AVCHD file that has been trimmed

PeterDuke

Newbie Joined: May 01, 2010 00:13 Messages: 7 Offline

Solved by

Oct 07, 2012 08:50

I normally have to trim off a small portion at the end of an AVCHD video clip because my camera does not stop recording immediately (perhaps it has to record to the end of the current GOP). If such trimmed clips are smart rendered, a frame near the end of the clip erroneously jumps forward, while the next frame does not advance (actually a bit more complicated than that) so that from then on, the frames are as they should be, to the end. I notice this effect very often, and it may always occur, but could go unnoticed if the scene is almost static where the jump would occur.

I took a clip that was 11 seconds and 24 frames (11:24) long (PAL system) and progressively trimmed one frame off the end. With trims at 11:23 down to 11:12, the jump occurred at 10:11, while for trims at 11:11 to 11:00 (as far as I tested), the jump occurred at 9:23. The location of the jump is presumably related to the last I-frame (picture) before the end.

If I turn off smart rendering, no jump occurs.

I placed a text subtitle near the start of the clip, forcing a non-smart render at that point, and no jump occurred towards the end.

Has anybody else noticed this effect? If not, would anyone try to reproduce it?

PeterDuke

Newbie Joined: May 01, 2010 00:13 Messages: 7 Offline

Oct 09, 2012 23:40

More information.

I wondered whether the source of the AVCHD file was significant. It was as shot from a Sony HDR-XR520 camera.

I converted the AVCHD file to MPEG2 HD using Sony Vegas and then converted this back to AVCHD using Corel Video Studio X5, Sony Vegas Pro 10, Cyberlink PowerDirector 11, Premiere Pro CS6 and Magix Video Pro X4.

I then trimmed the end of each output at the same place in PowerDirector 11 and smart rendered if possible. The last two outputs would not smart render. The first three had a jumped frame but the first one had its jump slightly later.

Thus the source of the AVCHD file can have a bearing on the result, but a problem still exists.

A jump is not easy to detect with some video content. What I used was a scene shot from a moving vehicle in which street light poles travel across the image at a steady rate.

This message was edited 1 time. Last update was at Oct 09. 2012 23:53

PatC

Senior Member Location: Suffolk UK Joined: Nov 17, 2009 14:00 Messages: 156 Offline

Oct 10, 2012 07:48

Hi PeterDuke

I have done extensive testing with AVCHD 1080i and 1080P from different Panasonic Camcorders
in PD9, PD10 and PD11. In all cases SVRT causes a glitch just after any transition between clips.
In PD9 and PD10 SVRT did reduce rendering time but it seems that, on my system at least,
PD11 with SVRT takes LONGER to encode than other options.
As PD is able to render video in a reasonable time without SVRT I do not use that option.
As I always tend to use the Fade Tansition in Overlap mode I have not recently tested if
clip to clip without transitions exhibits the glitch. From memory I think it did not but I will check
later today and post.

If without SVRT your results are good then I don't think you are missing much as I find that
PD is able to render results that look as good as the original even on HDTV and according to
MediaInfo have a very similar file structure.

I assume SVRT does work correctly for other file types that I do not have examples of,
but on my system Panasonic AVCHD definately causes a glitch.
I am using PAL 25/50fps video, 30/60fps may behave differently.

Patrick

PeterDuke

Newbie Joined: May 01, 2010 00:13 Messages: 7 Offline

Oct 10, 2012 21:17

Thanks for your reply. I will go through the motions of trying to get Cyberlink to accept this as a bug and hopefully something good might come out of it.