Log in

View Full Version : Suggestions for improving the quality!


Pages : 1 2 [3]

diimaan
3rd July 2010, 13:52
ohh goood!

how you spread across the sigma values?

from lower in sigma to higher in sigma4?

diimaan
3rd July 2010, 15:36
Hey, it showed some error in the second pass that it couldn't create file and some direct3d error?

could it be a problem cos i was playing football manager 2010! :D

Blue_MiSfit
3rd July 2010, 18:38
DOH! No playing video games while your PC is encoding!!!! :)

For the sigma values, I usually do a bell curve - kinda sorta...

sigma=.7, sigma2=1.5, sigma3=1.8, sigma4=.5 is a rough approximation of the shape I often go for. You need to spend time tweaking each band, ideally with the help of something like this

a=last
fft3dgpu(...)
interleave(a,last)

histogram(mode="luma")


It helps :devil:

Derek

diimaan
3rd July 2010, 22:22
ohh, thanks again Derek!
I found the problem, i think it's cos i was using the 64 bit x264 using pipebuf! :)

the 32 bit version works alright...
and yea fm 2010 is not gpu intensive, cos i used to play that and even i've played assassin'screed 2! but that time i did not used fft3dgpu

diimaan
3rd July 2010, 23:40
hey,

can anyone have a look at the samples??

samples (http://www.mediafire.com/?omohntgjtim)

i donno which is causing trouble in the test 1 file! some blocky tearing or artefacts appear... it happens to be the problem with the x264!

these are my avisynth script and x264


avisynth:
---------

Loadplugin("C:\Program Files (x86)\dgmpgdec158\DGDecode.dll")
Import("C:\Program Files (x86)\AviSynth 2.5\plugins\SeeSaw.avs")
Loadplugin("C:\Program Files (x86)\AviSynth 2.5\plugins\UnDot.dll")
Loadplugin("C:\Program Files (x86)\AviSynth 2.5\plugins\Decomb.dll")
Loadplugin("C:\Program Files (x86)\AviSynth 2.5\plugins\AutoCrop.dll")
Loadplugin("C:\Program Files (x86)\AviSynth 2.5\plugins\FFT3dGPU.dll")
Loadplugin("C:\Program Files (x86)\AviSynth 2.5\plugins\RemoveGrainSSE3.dll")

mpeg2source("C:\Users\Admin\Desktop\test\dvd2avi.d2v")
#Info()

AssumeTFF()
Telecide()
Decimate()

AutoCrop(mode=0, threshold=50)

UnDot()

a = last.RemoveGrain(mode=17)
b = a.FFT3dGPU(sigma=0.5, sigma2=1.5, sigma3=2, sigma4=0.5, plane=4, bw=48, bh=48, bt=4)
SeeSaw(a,b)



x264 for test 1:
---------------

@echo off
"D:\CLI\x264.exe" "C:\Users\Admin\Desktop\test\test.avs" --preset slower --sar 128:135 --crf 23 --output "C:\Users\Admin\Desktop\test\test.mkv"
pause


x264 for test 2:
---------------

@echo off
"D:\CLI\x264.exe" "C:\Users\Admin\Desktop\test\test.avs" --preset fast --me umh --sar 128:135 --crf 23 --output "C:\Users\Admin\Desktop\test\test.mkv"
pause



some setting in the slower profile did not like the file i think?
what do you ppl say?

edited:
also i tested the same source with veryslow and placebo presets! all goes fine! slower seems to be the problem!
strange... i'm using komisar's latest builds... is this a kind of any bug????

nm
4th July 2010, 10:54
hey,

can anyone have a look at the samples??

samples (http://www.mediafire.com/?omohntgjtim)

i donno which is causing trouble in the test 1 file! some blocky tearing or artefacts appear... it happens to be the problem with the x264!

I'm having a hard time spotting those artefacts. Which frames should I look at, or could you also post screenshots that exhibit the problem?

also i tested the same source with veryslow and placebo presets! all goes fine! slower seems to be the problem!
strange... i'm using komisar's latest builds... is this a kind of any bug????

Have you tried a second time with the slower preset? Maybe it was just a fluke in one of your filters (likely FFT3dGPU)?

diimaan
4th July 2010, 10:59
here are some screens...

http://i46.tinypic.com/2yjsyoi.jpg

http://i45.tinypic.com/fumdjp.jpg

http://i49.tinypic.com/eth3bc.jpg

i did tried without fft3dgpu! but still it's the same...

diimaan
4th July 2010, 11:00
screenshots are from test (1).mkv

nibus
4th July 2010, 11:06
DOH! No playing video games while your PC is encoding!!!! :)


Really? I play games all the time when I am encoding. Does it interfere with quality? Because I haven't noticed any issues with my encodes.

nm
4th July 2010, 11:50
screenshots are from test (1).mkv
I don't see those artifacts here, so it's a decoding problem. It looks like the weightp issue in older CoreAVC versions (http://forum.doom9.org/showthread.php?t=150776). Which decoder are you using?

diimaan
4th July 2010, 11:59
MPC video decoder? i think so!

or i should say haali splitter? :??

diimaan
4th July 2010, 12:08
I use media player classic home cinema...

and whatever decoder it uses, it should be the one... i'm sure i'm not using core avc... :)

also i've posted the source for this encode a few posts back, so you can test encode too...

here the link again (http://www.mediafire.com/?qj5m3mqllqn)


I can't see LoRd_MuldeR posting these days, hope he hasn't flew too South Africa! :D

No offense LoRd_MuldeR!!! :)

diimaan
4th July 2010, 15:21
here i'm yet again with a tricky source! :D

when i forced film and run a test encode it showed interlaced lines...
so i used honour pull down flags and i used telecide and decimate
still i can see a slight interlaced lines in the movie! :confused:

i'm uploading a sample here! what kind of source is this? and what should i do with it to encode!

i donno why i'm always end up with strange sources! :(

here you go friends (http://www.mediafire.com/?xzmck052umt)

Groucho2004
4th July 2010, 17:13
here i'm yet again with a tricky source! :D

This looks like some hybrid source.
You might consider reading through the Decomb tutorial that Donald was nice enough to supply and follow the steps to find the best approach for that source. You may actually learn something when you're not being spoon-fed the solutions every time.

diimaan
4th July 2010, 17:26
I would not ask for each and every source...
I do my test encodes with various parameters and update my post if i came to a solution!

I've read that tutorial and this one looks kinda like a 5:4 pull down! :o

diimaan
4th July 2010, 17:35
I think

TFM()
TDecimate()

should work fine!

hope am doing the right thing... :)

diimaan
4th July 2010, 20:35
well TFM() didn't do well on my source!

after testing with

yadif
tomsmocomp
nnedi3
tdeint

I'm settling with this


TDeint(mode=2, order=-1, field=-1, type=2, mtnmode=3, full=false, slow=2)


gives me great result for this particular source!!! :devil:
i wouldn't say it's fast... :p it averages around 8 fps with veryslow preset...

nm
4th July 2010, 22:33
I use media player classic home cinema...

and whatever decoder it uses, it should be the one... i'm sure i'm not using core avc... :)

Well, MPC-HC can be set to use internal libavcodec (which should decode your sample fine), some DirectShow decoder or hardware decoding through DXVA. You should try changing the decoder.

Also try VLC or MPlayer to confirm that the stream is ok.

diimaan
6th July 2010, 09:11
i'm running a two pass encode at 1500 kbps

but after the first pass, the log displays

x264 [info]: final ratefactor: 22.74

so if i run a crf with 22.74, will my output be of the same size compared to the one with a two pass encode?

nurbs
6th July 2010, 09:17
Generally no, because when you encode with crf you'll be using completely different settings than you used in the first pass.

diimaan
6th July 2010, 09:34
--preset placebo --tune film --sar 32:27 --pass 1 --bitrate 1500 --stats "encode.stats" --output NUL

this was my 1 pass...

plz don't pounce on me for placebo's :), i'm just trying to push my cpu... and am using nnedi3 on my avs... (i'm getting around 1.4 fps):p

so back to the topic

and if my crf is like this

--preset placebo --tune film --sar 32:27 --crf 22.74 --output "*.mkv"


still my outputs will be of different sizes?

J_Darnley
6th July 2010, 10:00
Probably not different sizes but there will be lower quality in the first one because that is 1-pass ABR. Running the second pass will correct that.

diimaan
6th July 2010, 10:18
Probably not different sizes but there will be lower quality in the first one because that is 1-pass ABR. Running the second pass will correct that.

Oh yeah, thanks darnley...

that was the first pass only i did run a 2nd pass with mkv output! :)


--preset placebo --tune film --sar 32:27 --pass 1 --bitrate 1500 --stats "encode.stats" --output NUL
--preset placebo --tune film --sar 32:27 --pass 2 --bitrate 1500 --stats "encode.stats" --output "*.mkv"


this is the full code... :)