View Full Version : JVC 24p Conversion
tuprware
20th March 2008, 00:19
I have been trying to use avisynth and virtual dub to convert a file captured from a JVC HD100 shot at 24p. I captured with hdvsplit. The file info in VD shows as 1280x720@59.94fps length is 1:39.54.
So I captured about 31 seconds, so the length of the clip is multiplied. This site provided me info on how JVC does their pattern:
http://digitalcontentproducer.com/videoencodvd/revfeat/hdv_102405/
So I have the pattern and I've tried: pulldown, SelectEvery, telecide and decimate. Not much luck.
You can download the clip I was testing with at:
http://www.tupware.com/videos/24p.m2t
Any help would be appreciated.
thanks,
tup
scharfis_brain
20th March 2008, 00:20
the link to your sample video does not work.
Blue_MiSfit
20th March 2008, 00:33
Ditto.
Try just TFM().TDecimate() :)
~MiSfit
scharfis_brain
20th March 2008, 01:21
JVC seems to use RFF flags with MPEG2 720p59.94 video.
I'd suggest to use "ignore Pulldown" in DGIndex and then assumefps(23.976) in the script.
That would avoid any IVTC-process.
Guest
20th March 2008, 01:52
720P RFFs are usually frame repeats, and sometimes irregular. We really need to see the stream to make a correct recommendation.
2Bdecided
20th March 2008, 13:42
The stream downloaded fine here.
Ignore pulldown flags and assumefps(24) seem to work fine - certainly on the part where the car moves, it moves correctly. There's not much motion on the rest of the clip to judge.
Cheers,
David.
scharfis_brain
20th March 2008, 16:17
@2Bdecided: assumefps(23.976) has to be taken if ignore pulldown flags is used. assumefps(24) will introduce an increasing A/V offset.
also I had success with honor pulldown flags and tdecimate(mode=2, rate=23.976)
further selecteven().tdecimate() works, too.
But I recommend testing to ignorethe flags on a full length stream first. (Check the duration. If it fits to the 'hnonored' video, it'll be all right)
@neuron2: is it possbile to extend forced film to 720p60 MPEG2?
@tuprware: try to disable the internal denoiser of the camera. There are weird thresholding artifacts present in the video.
2Bdecided
20th March 2008, 17:21
@2Bdecided: assumefps(23.976) has to be taken if ignore pulldown flags is used. assumefps(24) will introduce an increasing A/V offset.Sorry - typo - I actually used 25 to integer match my monitor's 75Hz rate to best judge smooth motion. Of course you must use correct rate for final processing, and any irregular pattern could still throw it off sync.
(I know you know that - just emphasising!).
Cheers,
David.
tuprware
20th March 2008, 18:08
Using a d2v with tdecimate(mode=2, rate=23.976), instead of directshow source worked well for getting the clip length and frame rate correct. But it introduced interlacing and used the wrong decompressor (my machine chose divx and YV12-maybe the cause of noise too). I've been using directshow source to open the m2t file and that uses YUY2 as the decompressor and makes the images look a lot better.
Thanks for all the help, I've been trying to get this to work all week and in just a few hours its way closer than I've been in days.
Guest
20th March 2008, 18:51
is it possbile to extend forced film to 720p60 MPEG2? Yes. It's already on my internal to-do list. It will rise in priority if I ever need to process that kind of material.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.