stattik_transcode 0.3.1 released

[-stattik-]

What is your source material? Perhaps the script is guessing it incorrectly. Thanks for the feedback.

[-stattik-]

Version 0.3.1a of the script has been released.

Here's the release notes:

0.3.1a
* Fixed a bug that broke the queuing of files after the initial pass with Automated Directory Transcoding enabled
* Renamed video_fps to output_video_fps to address any confusion
* Updated the DVP-642 support. It's been tested on interlaced and HD content(1080i). Progressive content is being worked on. It's a work in progress and any feedback would be appreciated.


I would recommend anyone using the Automated directory transcode or DVP-642 profiles to use this version as it addresses some issues.

[chud]

my understanding is that with the use of "-vf filmdint" you have to set the output to 24fps.

this is quite an impressive piece of work stattik. i have by own elaborate scheme for transcoding video to xvid, but it is no where as finely polished as this.

when you first introduced this tool i was keen to see your mencoder settings, specifically in your treatment of deinterlacing. i played around with many settings and none were terribly rewarding. i have done many back to back transcoding comparisons and have found dropping to 24fps to be too jittery for me. i have settled on the lowpass5 video filter for deinterlacing. i am not sure if you have checked it out.

of course, with 30fps in your final video, you have less bits per frame to work with for the same bitrate, but i used the same bitrate as the 24fps encodings and preferred the PSNR tradeoff for smooth video.

m.

[-stattik-]

Settings are always being tweaked. The difference in input files make it difficult to find a solution for everything. filmdeint may be phased out in favor of pp=md. Mencoder seems to have no lack if deinterlacers. I haven't tested lowpass5 yet but since it works for you, perhaps I'll take a look at it. Thanks for the info.

[UncleJohnsBand]

What is your source material? Perhaps the script is guessing it incorrectly. Thanks for the feedback.

Source comes from PVR150 or PVR500 recording mpeg @ 29.97 fps.

[UncleJohnsBand]

my understanding is that with the use of "-vf filmdint" you have to set the output to 24fps.

I wonder if that is why I see a ton of duplicate frames being dropped when I up the fps to 30000/1001.

[wtg]

I wonder if that is why I see a ton of duplicate frames being dropped when I up the fps to 30000/1001.

I don't think that's the reason because I encode mine at at 24fps and gets lots of duplicates dropped too.

[chud]

I wonder if that is why I see a ton of duplicate frames being dropped when I up the fps to 30000/1001.

that has been my experience with it. there can be significant duplicate frames when doing a filmdint 24fps encode, but if you force 30fps it is a constant stream.

try replacing the "-vf filmdint" with "-vf pp=de/l5" and leave the fps at 30. it is all a bit subjective, but if you have noticed the jittery pans you may find it an improvement...

[-stattik-]

Thanks for all the feedback.

Here's a test loop you can run. Drop a file on here and it will create six files using each of the filters in this thread and a few more. The six filenames show the filter used, FPS, and GMC(if GMC enabled). The output window stays up to you can check PSNR (Peak Signal to Niose Ratio.) Try it with interlaced/progressive content and see what works best for you.

Thanks.

[UncleJohnsBand]

Thanks for all the feedback.

Here's a test loop you can run. Drop a file on here and it will create six files using each of the filters in this thread and a few more. The six filenames show the filter used, FPS, and GMC(if GMC enabled). The output window stays up to you can check PSNR (Peak Signal to Niose Ratio.) Try it with interlaced/progressive content and see what works best for you.

Thanks.

Nice little script...very helpful in determining what works best....

I used it on 6 or so recordings (all from my hauppage tunners) but recorded from different channels.

No filter and the pp=de-l5 wer the worst PSNR filters....and they looked the worst. There PSNR where very close to each other.

filmdint and pullup were next....again both had very close PSNR but both dropped a lot of frames while transcoding.

The best two were yadif and pp=md. They consistently had the best PSNR and were very close to each other with very little fame droppage.

Even though yadif had a slightly higher PSNR than pp=md I like how pp=md renders....it seems to be a little softer in appearance.

I have a question......my recordings are in 720x480 format....your script is defaulted to 512x384.....is there a particular reason for that size? I set the script to be 720x480 but thought I would ask if there was any benefit in reducing the output size.