3:2 pulldown on 29.970i Anime + addtional AVS script configuration

[Guest]

A third pass is not necessary. Just stick with two passes.

[TheProfosist]

ok will do but ive always used automated 3 pass mode so now i have to set each pass individually correct? and from what ive been reading the setting on the 2 passes should be as close together as possible. so i plan to set both passes as close to my 3 pass settings as possible. would you happen to know what the x246 syntax is for setting the maximum bit rate? right non im using --vbv-maxrate 9800 for DVDs and --vbv-maxrate 40000 for BDs but the setting doesnt seem to do anything at all the max bitrates after encoding DVDs are sometimes 10Mbps+

[Guest]

Settings help for x264 is readily available. Use Google to find it.

Anyway, it's off topic for this thread.

[TheProfosist]

Settings help for x264 is readily available. Use Google to find it.

Anyway, it's off topic for this thread.

i know and ive read and etc wend through all x264 settings on its wiki but the best i could come up with was the vbv-maxbitrate but that doesnt seem to do anything for me. it might only affect cbr and the only other max that can be set is qp but that deals with quantizer. i encode abr

i shouldnt have any major questions for the time being as i will be finally starting the encode for moribito. i will encode the first episode and check the result hopefully everythings good

[Guest]

i know and ive read and etc wend through all x264 settings on its wiki but the best i could come up with was the vbv-maxbitrate but that doesnt seem to do anything for me. That's because you don't know what a VBV model is and how it works. You're dabbling. You must have a firm technical foundation for the things you are trying to use.

[TheProfosist]

That's because you don't know what a VBV model is and how it works. You're dabbling. You must have a firm technical foundation for the things you are trying to use.

well i was reading up i know what its for its to restrict bitrate and buffer so streams will be compatible with certain devices. but it should work to restrict streams to original maximum bitrates.

i try to do the research the best i can before asking.

[Guest]

but i should work to restrict streams to original maximum bitrates. Why "should" you do that? And do you know that is not the purpose/effect of implementing VBV?

[TheProfosist]

*"it should" sorry the 't' there was kinda important

because i shouldnt have a bitrate spike exceeding the maximum bit rate of the source. it not really all that big of a deal i just though it odd to have a max bitrate of 10.2mbps after encoding when then max bitrtate of the source is less than 9mbps

[Guest]

You're not getting it!

To expose your lack of understanding, I'll ask you: What is your definition of max bitrate?

[TheProfosist]

You're not getting it!

To expose your lack of understanding, I'll ask you: What is your definition of max bitrate?

the maximum bitrate the encoding peaks at at any one frame just like what DGindexNV reports as max bitrate

[Guest]

That's not what DGIndexNV reports.

[TheProfosist]

That's not what DGIndexNV reports.

then what kind of max bitrate does it report? are there different kinds.

what i think of as max bitrate is when i play it back with a bitrate monitor on the highest bitrate i see is the max bitrate to me

[TheProfosist]

well heres a link to 100mb of the encoded first episode http://www.mediafire.com/file/ocwzdnnzuuz/Moribito 01-001.mkv

[Guest]

Suppose at one moment of time I hand you the entire stream. What is the maximum bitrate according to your understanding?

[TheProfosist]

Suppose at one moment of time I hand you the entire stream. What is the maximum bitrate according to your understanding?

now im just all confused with the whole max bitrate subject

well take a look at that file when you have time.
the timecodes file says its 99% film could i instead then just pulldown to 23.976. the only problem i have in doing this is that DGindex file says its 68.62% FILM

[Guest]

now im just all confused with the whole max bitrate subject Yes, the idea of max (peak) bitrate is meaningless unless the period over which the transfer takes place is specified. The max bitrate measured over one second may be quite different from that measured over one minute, etc. So the notion of max bitrate is less useful than most suppose. You need to understand the VBV model to understand the factors that you should pay attention to.

well take a look at that file when you have time.
the timecodes file says its 99% film could i instead then just pulldown to 23.976. the only problem i have in doing this is that DGindex file says its 68.62% FILM I'd be inclined to trust DGIndex. But can't you inspect the video as I explained several times to see what you really have?

BTW, I almost never look at people's encode samples. I'm interested only in source files and corresponding procedures.

[TheProfosist]

i removed the vbv stuff from the script since i shouldnt need it


ill do my best to try to follow whats on that FAQ page that should tell me which type of it is.


during the opening and high motion scenes i get aaa bb ccc dd which is 3:2
but during everything else i get aaaaa bbbbb ccccc this is not listed on your faq my guess is that this is most likely got recognized as 3:2 and thats why the timecodes file said 99% film but thats just a guess u would probably know better than i

[osgZach]

Assuming those "aaaaa bbbbb ccccc" have no visible interlacing/combing it sounds like you're describing a 30fps progressive section of video.

But you haven't mentioned why you are bringing it up.. are you just curious, or is playback jerky from possible duplicate frames, or some problem like that? Otherwise if its playing fine. Don't worry about it. A VFR file is -supposed- to have different sections running at different frame rates, which means you will see different patterns.

Or if I missed the reason, humor me.. I haven't read this thread for a while and jumped to the last page.

[Dogway]

I would like to know how to do pulldown removal as well, I have 2 blended frames for 3 progressive. Source is 29.970i and I would like to convert it to 24p. Thing is I have tried everything I saw on the forums but it doesnt work. The deinterlacers dont work as expected (they leave ugly strtipes) and Decimate doesnt remove the blended frames. I post the source.

[Deleted]

The code that has worked best but still not perfect is this:

Telecide(guide=1,back=1)
Vinverse()
SRestore(FRate=24)

[The Scientist]

I post the source.The source is an AVI?

[Dogway]

Yes, Its the best I could get. Its a part of a videoclip. Is it doable?

[Guest]

Yes, Its the best I could get. Its a part of a videoclip. Is it doable? We can't help you with that material per rule 6. Do not post about such material again.

[Dogway]

Thanks. I will post on the offtopic or whatever, its just I dont understand why should I post DVD content, which actually is the end point (Original DVD last 3 more times than any other storage medium), and not things I got through other mediums because it's hard as hell to get a DVD copy. (rare stuff). I mean where do you buy short movies? videoclips? TV bumpers? (...). Thanks, anyway I understand your posture.

[Motenai Yoda]

dgindex i think reports a correct value for the video stored.
but with tfm or telecide u do a inverse telecine, so the stream after one of these filters may be very different.
In this case dgindex says 65% film but the video probably has been strangely telecined, but isn't video vfr.
Moreover tdecimate has a threshold to determinate a duplicate.


for tdecimate 2 pass I think u shoud do the first pass without encoding, eg with virtualdub, preview input.
and so make the second pass with a lossless codec eg megui+memcoder or virtualdub+lagarith.
Hence encode.

sorry for my English

[Guest]

dgindex i think reports a correct value for the video stored Who are you responding to?

[Motenai Yoda]

to this
I'd be inclined to trust DGIndex. But can't you inspect the video as I explained several times to see what you really have?
most of hard telecined anime was recognized as video by DGindex, but often are only a 24p telecided, aren't vfr.
also most of progressive video was recognized as interlaced.

[Guest]

also most of progressive video was recognized as interlaced. Here you make the common mistake of confusing the nature of the content with the method of encoding. DGIndex reports only whether the stream was *encoded* interlaced or progressive. That is NOT the same thing as reporting the content as interlaced or progressive. I've explained this so many times (not to mention the clear statement in the documentation) that I'm tempted to vow never again to waste bandwidth saying it yet again.

[Motenai Yoda]

yes it's a simil thing, imho DGindex use stream or encoding flag to differentiate hard pulldown or ntsc-video from soft pulldown or ntsc-film, it doesn't make a field analysis.

[osgZach]

Wait... are you attempting to explain to Neuron2 how his own program works? lol :p

Uh.. anyway.. small suggestion. Maybe changing the label to something like "Encoded As" might help reduce the annoying misunderstanding.. Or then again people are people and this is the Internet, so maybe not... Oh well

[Motenai Yoda]

Wait... are you attempting to explain to Neuron2 how his own program works? lol
No, I didn't want to do this, but explain my point of view. I haven't read the source, so can't know how it works, these are my assumptions about it.
neuron2 and his program's are a fundamental and important, are great! but some things are over the idea, and the possibilities of DGindex.

[TheProfosist]

say if i wanted to use crf instead of 2 pass? would it still be beneficial to run 2 passes for vfr instead of using the single pass vfr? the reason why is i have been thinking of switching from 2pass mode to crf mode

examples fo vfr code
-------------------------------------------------------------

~~~~~~~~~~
1 Pass VFR
~~~~~~~~~~

TFM(order=1)
TDecimate(mode=3, hybrid=2)


-------------------------------------------------------------

~~~~~~~~~~
2 Pass VFR
~~~~~~~~~~

1st Pass
--------

TFM(order=1, slow=2, output="C:\TEMP\matches.txt")
TDecimate(mode=4, output="C:\TEMP\metrics.txt")


2nd Pass
--------

TFM(order=1, slow=2, input="C:\TEMP\matches.txt")
TDecimate(mode=5, hybrid=2, input="C:\TEMP\metrics.txt", tfmIn="C:\TEMP\matches.txt", mkvOut="C:\TEMP\mkv-timecodesfile.txt")

[TheProfosist]

im gessing 2 pass wouldnt work with crf because the stats (metrics, matches) wouldnt match up. so can i still specify where i want the time codes file to got out like in 2 pass?

[osgZach]

As far as I know, single pass VFR w/timecodes should be the same as with 2-pass, just supposedly takes longer since its calculating everything at once.

Or I could be wrong. Sure someone will correct me if so.

[TheProfosist]

idk if this would work but would i be able to use the queue analysis pass to do the first pass of the vfr then do the encode using crf?

[TheProfosist]

also for an avs script what is the proper order for the scripts to be in right now i have

crop
deinterlace/telecene
denoise/other filters
resize

does this seem correct because MeGUI has it in this order

deinterlace/telecene
crop
resize
denoise

[osgZach]

I don't use Analysis pass much, so I can't really comment.

As for your filter order, I usually do cropping and resizing last. Sometimes it comes down to personal preference.

[kypec]

I would definitely put cropping AFTER deinterlace, otherwsie you might end up with screwed video.

[TheProfosist]

I would definitely put cropping AFTER deinterlace, otherwsie you might end up with screwed video.

why? cropping the picture shouldnt affect the interlace at all. i mean unless you can show specific reason why errors would happen or what cropping would do to the interlace i usually like reasons behind doing things.

[kypec]

Because I learned myself that it's usually safer to deinterlace first and only then do other stuff like cropping or resizing. Read this thread (http://forum.doom9.org/showthread.php?t=149820&highlight=crop+deinterlace) for advices of more experienced users. Cropping removes irreversibly some parts of frame and one must remember that interlaced content makes complete frame of two consecutive fields with even/odd lines only. If you crop few lines of even field but keep complementary lines in odd field then deinterlacer will most likely produce garbage. I observed such situations when encoding episodes of Friends DVD myself. During frame stepping (Yadif used after cropping) there were occasionally sort of "ghost" frames around scene transitions. Not very noticable during normal playback but still...why do something in a wrong way when there is better way ;)

[TheProfosist]

Because I learned myself that it's usually safer to deinterlace first and only then do other stuff like cropping or resizing. Read this thread (http://forum.doom9.org/showthread.php?t=149820&highlight=crop+deinterlace) for advices of more experienced users. Cropping removes irreversibly some parts of frame and one must remember that interlaced content makes complete frame of two consecutive fields with even/odd lines only. If you crop few lines of even field but keep complementary lines in odd field then deinterlacer will most likely produce garbage. I observed such situations when encoding episodes of Friends DVD myself. During frame stepping (Yadif used after cropping) there were occasionally sort of "ghost" frames around scene transitions. Not very noticable during normal playback but still...why do something in a wrong way when there is better way ;)

thanks much for the info i like to know this stuff incase i ever have to look into something deeper or have to help someone else. i personally hate throwing out info without reasons behind them it just doesnt seem very credible

so new order

deinterlace/reverse pulldown
crop
denoise/sharpen/etc
resize

[TheProfosist]

i looked into crf some more and decided to go with it because it saves space even though I am not worried about it why not save space where you can i have also learned that u can use the analysis pass to do the first pass of 2 pass VFR. i have also found a more accurate way of decimation using 2 passes getting the metrics and matches on the first pass