Avisynth Script for doubling video frame rates - Page 40 - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #1171 of 1968 Old 03-22-2011, 11:00 AM
Newbie
 
ayawlex's Avatar
 
Join Date: Jan 2011
Posts: 8
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by widezu69 View Post

I guessed I rushed into it too fast. Thanks for all the info guys! Will test further!
Edit: Looks good with hardly any performance hit. I took Sub's InterFrame and translated the data into one usable script that I using now:

SetMTMode(1,16)
V = ffdshow_source()
SetMTMode(2)
G = V.RemoveGrain(mode=1)
H = G.MSuper(pel=2, hpad=8, vpad=8, rfilter=4)
S = V.MSuper(pel=2, hpad=8, vpad=8, rfilter=4)
A = H.MAnalyse(isb=true, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
B = H.MAnalyse(isb=false, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
C = S.MRecalculate(A, blksize=8, blksizev=8, search=3, searchparam=1)
D = S.MRecalculate(B, blksize=8, blksizev=8, search=3, searchparam=1)
E = S.MRecalculate(C, blksize=4, blksizev=4, search=3, searchparam=1)
F = S.MRecalculate(D, blksize=4, blksizev=4, search=3, searchparam=1)
V.MFlowFps(S, E, F, num=FramerateNumerator(V)*2, den=FramerateDenominator(V)*1, mask=2, ml=10000, blend=false)
GetMTMode(false) > 0 ? distributor() : last

Have to dct=10 though because 5 has a pretty big performance hit for me.
Credit goes to SubJunk.

gives me an error "no function named RemoveGrain"
ayawlex is offline  
Sponsored Links
Advertisement
 
post #1172 of 1968 Old 03-22-2011, 12:29 PM
Senior Member
 
widezu69's Avatar
 
Join Date: Feb 2010
Posts: 319
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
You need to download RemoveGrain from here:
http://avisynth.org/mediawiki/Removegrain
Get the latest beta and use RemoveGrainSSE3.dll if you have an Intel Processor. Unsure which one to use if you have an AMD one.
widezu69 is offline  
post #1173 of 1968 Old 03-22-2011, 01:58 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Nice one widezu. I think the latest version of RemoveGrain is the pre-release, not the beta. That comes with a few variations for different CPUs, too.
SubJunk is offline  
post #1174 of 1968 Old 03-22-2011, 05:04 PM
Senior Member
 
widezu69's Avatar
 
Join Date: Feb 2010
Posts: 319
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
You know what you are absolutely right. The pre-release version is pretty quick! With dct=10 I have about 70-80% usage most times on 720p playback. I am on 2.8GHz i7 on all 8 threads.
Been doing some research into RemoveGrain and I'm using mode=3. 2 is default but the effects are too subtle, 3 does a bit more and cleans up your picture so it obviously helps with interpolation. So to implement just use the above script and change mode to 3.
Edit: I've examined the frames and it turns out that the Grainless images are not being rendered and are only used for the interpolation vector creation. So you can have the highest level of blurriness for interpolation and your rendered image is still as sharp as before!
widezu69 is offline  
post #1175 of 1968 Old 03-22-2011, 10:41 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by widezu69 View Post

I've examined the frames and it turns out that the Grainless images are not being rendered and are only used for the interpolation vector creation. So you can have the highest level of blurriness for interpolation and your rendered image is still as sharp as before!

I probably should have pointed that out explicitly; I wouldn't have added it otherwise I see why you were hesitant to add it now! Didee thought the same thing so you're in good company at least
SubJunk is offline  
post #1176 of 1968 Old 03-23-2011, 01:46 AM
Senior Member
 
widezu69's Avatar
 
Join Date: Feb 2010
Posts: 319
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Update: Here is a new version and the rendered images are all pre-filtered. Notice I got rid of one of the MSuper lines completely:

SetMTMode(1,16)
V = ffdshow_source()
G = V.RemoveGrain(mode=2)
SetMTMode(2)
S = G.MSuper(pel=2, hpad=8, vpad=8, rfilter=4)
A = S.MAnalyse(isb=true, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
B = S.MAnalyse(isb=false, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
C = S.MRecalculate(A, blksize=8, blksizev=8, search=3, searchparam=1)
D = S.MRecalculate(B, blksize=8, blksizev=8, search=3, searchparam=1)
E = S.MRecalculate(C, blksize=4, blksizev=4, search=3, searchparam=1)
F = S.MRecalculate(D, blksize=4, blksizev=4, search=3, searchparam=1)
G.MFlowFps(S, E, F, num=FramerateNumerator(V)*2, den=FramerateDenominator(V)*1, mask=2, ml=10000, blend=false)
GetMTMode(false) > 0 ? distributor() : last
widezu69 is offline  
post #1177 of 1968 Old 03-23-2011, 06:43 AM
Newbie
 
hartman75's Avatar
 
Join Date: Mar 2011
Posts: 1
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Greetings for all!

Need some help / advice:
I get a lot of crashes in mpc , (something like "APPCRASH: "mpc-hc.exe", error module: "mvtools2.dll" ver2.5.11.1 etc...)...it always occurs when i´m (jumping) seeking the movie with a mouse.

Are there any ideas what´s wrong?...maybe a script issue?

My Specs:
-core i7 920@3.6/6GB, Win7x64, X-fi (optical audio out)
-AMD T1055@3.8 / 4GB , Win7x64 (hdmi audio out)

Movies are mostly 720p/1080p MKV´s.

I´ve installed all the software with according the instructions posted in this thread and I´m using this script with my i7:

-----
SetMemoryMax(512)
SetMTMode(2,16)
ffdShow_source()
SetMTMode(2)
super = MSuper(pel=2, hpad=16, vpad=16, rfilter=4)
backward_1 = MAnalyse(super,chroma=false,isb=true, blksize=16, searchparam=1, search=3,plevel=0, sadx264=7,dct=5)
forward_1 = MAnalyse(super,chroma=false, isb=false, blksize=16, searchparam=1, search=3, plevel=0, sadx264=7,dct=5)
backward_2 = MRecalculate(super, backward_1, chroma=false, blksize=8, searchparam=0, search=3, sadx264=7)
forward_2 = MRecalculate(super, forward_1, chroma=false,blksize=8, searchparam=0, search=3, sadx264=7)
Framerate >=23 && Framerate <= 24 ? MBlockFps(super, backward_2, forward_2, num=60, den=1, mode=0) : last
GetMTMode(false) > 0 ? distributor() : last
-----

and yes, I´m a 60fps addict
hartman75 is offline  
post #1178 of 1968 Old 03-23-2011, 08:10 AM
Member
 
travolter's Avatar
 
Join Date: May 2009
Posts: 162
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
@hartman75

things to try:
- use other mvtools and avisynth. check previous posts to get some stable versions.
- increase setmemorymax(1024)

@8:13

I tested the scripts.. very low cpu usage but I need more fluidity. I continue in love with this one:
http://www.avsforum.com/avs-vb/showp...postcount=1161

If you have enough cpu power.. try it. Max fluidity to the eye
travolter is offline  
post #1179 of 1968 Old 03-23-2011, 11:31 AM
Senior Member
 
widezu69's Avatar
 
Join Date: Feb 2010
Posts: 319
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
@hartman you may have a multi threading problem. I'd recommend you get this:
http://www.mediafire.com/?rrt2fcmrdjzw0rc
And stick it in your system32 folder or SysWow64 (if you have 64 bit). It should solve your seeking issues.
widezu69 is offline  
post #1180 of 1968 Old 03-23-2011, 12:33 PM
Member
 
mark007's Avatar
 
Join Date: Dec 2006
Posts: 196
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Hi SubJunk, 8:13. Are these newer mvtools dlls you have posted recently compiled differently or what? Are they faster? I'm currently now using the one from SubJunks zip. I see a new mvtools version was released v2.5.11.2, but the changes don't seem relevant to the scripts posted here.

2.5.11.2 beta (20.03.2011 by Fizick)
MCompensate: fixed crash with too small npad, vpad (thanks to '-Vit-').
mark007 is offline  
post #1181 of 1968 Old 03-23-2011, 02:01 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by mark007 View Post

Hi SubJunk, 8:13. Are these newer mvtools dlls you have posted recently compiled differently or what? Are they faster? I'm currently now using the one from SubJunks zip. I see a new mvtools version was released v2.5.11.2, but the changes don't seem relevant to the scripts posted here.

2.5.11.2 beta (20.03.2011 by Fizick)
MCompensate: fixed crash with too small npad, vpad (thanks to '-Vit-').

Hi Mark
The one in my zip is that new beta, if you see the guide changelog it is listed there
I use the official builds since the custom ones I have tested have been slower. I haven't tried 8:13's one.

Quote:
Originally Posted by widezu69 View Post

Update: Here is a new version and the rendered images are all pre-filtered. Notice I got rid of one of the MSuper lines completely:

SetMTMode(1,16)
V = ffdshow_source()
G = V.RemoveGrain(mode=2)
SetMTMode(2)
S = G.MSuper(pel=2, hpad=8, vpad=8, rfilter=4)
A = S.MAnalyse(isb=true, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
B = S.MAnalyse(isb=false, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
C = S.MRecalculate(A, blksize=8, blksizev=8, search=3, searchparam=1)
D = S.MRecalculate(B, blksize=8, blksizev=8, search=3, searchparam=1)
E = S.MRecalculate(C, blksize=4, blksizev=4, search=3, searchparam=1)
F = S.MRecalculate(D, blksize=4, blksizev=4, search=3, searchparam=1)
G.MFlowFps(S, E, F, num=FramerateNumerator(V)*2, den=FramerateDenominator(V)*1, mask=2, ml=10000, blend=false)
GetMTMode(false) > 0 ? distributor() : last

Mode 2 in RemoveGrain is the default mode Also, so is pel=2, hpad=8 and vpad=8 in msuper so you can remove all those things if you want
I'm encoding some test clips to compare the removegrain change you did now. I'm hesitant to not include the grain at all, since I think it makes logical sense to use it in recalculate, but maybe I'm wrong and we'll see.

Quote:
Originally Posted by 8:13 View Post

Look at this code: setmtmode(2, 2), the 2 on the right is supposed to match the number of cpu cores your computer processor has;
My Intel E6600 has 2 cores so I set it to 2.

Or just set it to 0 which means automatic detection

Quote:
Originally Posted by 8:13 View Post

Subjunk, it you compare the blendfps version #2, to your function interframe then please judge 20 frames in sequence as only cherry picking on frame from many is not fair imvho.

I usually test sequences of at least 50 interpolated frames in a row and I use a numbered list and tick which frame is better each time, then the one with the most better frames wins.
I repeat this for several videos (although lately I have just been using PetitDragon's test video for convenience) and add up the total scores and based on that result I come back and post which is better with a few screenshots that are indicative of the trend in the test
However your scripts haven't done very well in that test so I have stopped testing them.
If you think you have made a particular leap in quality, I can test if you want

Quote:
Originally Posted by hartman75 View Post

Greetings for all!

Need some help / advice:
I get a lot of crashes in mpc , (something like "APPCRASH: "mpc-hc.exe", error module: "mvtools2.dll" ver2.5.11.1 etc...)...it always occurs when i´m (jumping) seeking the movie with a mouse.

Are there any ideas what´s wrong?...maybe a script issue?

My Specs:
-core i7 920@3.6/6GB, Win7x64, X-fi (optical audio out)
-AMD T1055@3.8 / 4GB , Win7x64 (hdmi audio out)

Movies are mostly 720p/1080p MKV´s.

I´ve installed all the software with according the instructions posted in this thread and I´m using this script with my i7:

-----
SetMemoryMax(512)
SetMTMode(2,16)
ffdShow_source()
SetMTMode(2)
super = MSuper(pel=2, hpad=16, vpad=16, rfilter=4)
backward_1 = MAnalyse(super,chroma=false,isb=true, blksize=16, searchparam=1, search=3,plevel=0, sadx264=7,dct=5)
forward_1 = MAnalyse(super,chroma=false, isb=false, blksize=16, searchparam=1, search=3, plevel=0, sadx264=7,dct=5)
backward_2 = MRecalculate(super, backward_1, chroma=false, blksize=8, searchparam=0, search=3, sadx264=7)
forward_2 = MRecalculate(super, forward_1, chroma=false,blksize=8, searchparam=0, search=3, sadx264=7)
Framerate >=23 && Framerate <= 24 ? MBlockFps(super, backward_2, forward_2, num=60, den=1, mode=0) : last
GetMTMode(false) > 0 ? distributor() : last
-----

and yes, I´m a 60fps addict

Try changing the second line to SetMTMode(2,0) and experiment with different values for SetMemoryMax. Try 768 and 1024.
SubJunk is offline  
post #1182 of 1968 Old 03-23-2011, 02:28 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by SubJunk View Post

I'm encoding some test clips to compare the removegrain change you did now. I'm hesitant to not include the grain at all, since I think it makes logical sense to use it in recalculate, but maybe I'm wrong and we'll see.

After testing it seems that while there isn't much quality loss, there are subtle details lost by doing it your way and it doesn't seem to have any benefits for motion analysis, so I think it's best to use 2 super clips like mine does.
It is a common practice when denoising with MVTools2, so it makes sense to do it here too
SubJunk is offline  
post #1183 of 1968 Old 03-23-2011, 02:52 PM
Senior Member
 
widezu69's Avatar
 
Join Date: Feb 2010
Posts: 319
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Ok thanks for the heads up! Here's what I have:

SetMTMode(1,16)
V = ffdshow_source()
G = V.RemoveGrain(mode=17)
SetMTMode(2)
S = V.MSuper(pel=2, hpad=8, vpad=8, rfilter=4)
H = G.MSuper(pel=2, hpad=8, vpad=8, rfilter=4)
A = H.MAnalyse(isb=true, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
B = H.MAnalyse(isb=false, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
C = H.MRecalculate(A, blksize=8, blksizev=8, search=3, searchparam=1)
D = H.MRecalculate(B, blksize=8, blksizev=8, search=3, searchparam=1)
E = H.MRecalculate(C, blksize=4, blksizev=4, search=3, searchparam=1)
F = H.MRecalculate(D, blksize=4, blksizev=4, search=3, searchparam=1)
V.MFlowFps(S, E, F, num=FramerateNumerator(V)*2, den=FramerateDenominator(V)*1, mask=2, ml=10000, blend=false)
GetMTMode(false) > 0 ? distributor() : last

I am using the grainless data in all of the passes. Only when MVTools draws the frames does it use the original sharp data. Looking around the general consensus seems that mode=17 is a pretty decent denoiser so I'm sticking to that. Any thoughts?
widezu69 is offline  
post #1184 of 1968 Old 03-23-2011, 03:44 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by widezu69 View Post

Ok thanks for the heads up! Here's what I have:

SetMTMode(1,16)
V = ffdshow_source()
G = V.RemoveGrain(mode=17)
SetMTMode(2)
S = V.MSuper(pel=2, hpad=8, vpad=8, rfilter=4)
H = G.MSuper(pel=2, hpad=8, vpad=8, rfilter=4)
A = H.MAnalyse(isb=true, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
B = H.MAnalyse(isb=false, blksize=16, plevel=0, search=3, searchparam=3, badrange=(-24), dct=10)
C = H.MRecalculate(A, blksize=8, blksizev=8, search=3, searchparam=1)
D = H.MRecalculate(B, blksize=8, blksizev=8, search=3, searchparam=1)
E = H.MRecalculate(C, blksize=4, blksizev=4, search=3, searchparam=1)
F = H.MRecalculate(D, blksize=4, blksizev=4, search=3, searchparam=1)
V.MFlowFps(S, E, F, num=FramerateNumerator(V)*2, den=FramerateDenominator(V)*1, mask=2, ml=10000, blend=false)
GetMTMode(false) > 0 ? distributor() : last

I am using the grainless data in all of the passes. Only when MVTools draws the frames does it use the original sharp data. Looking around the general consensus seems that mode=17 is a pretty decent denoiser so I'm sticking to that. Any thoughts?

Did you find that even thSCD1=350 was too low? I have noticed that 300 makes it ignore some frames but I also like the clean scene-changes. Maybe we can try setting using thSCD1=300 but also raising thSCD2. It defaults to 130 so maybe 160 or 180 will make it more accurate. Just an idea anyway.

I currently have some tests running on which mode of RemoveGrain is best for this purpose and I'll post back when that's done.
SubJunk is offline  
post #1185 of 1968 Old 03-23-2011, 05:03 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
It doesn't seem to matter much but mode=17 seems to have a slight edge in most scenes.

Edit: OK I've just updated InterFrame to 1.2 which adds some more presets that will make it more suitable for realtime playback. See the link in my signature for details
SubJunk is offline  
post #1186 of 1968 Old 03-24-2011, 03:00 AM
AVS Special Member
 
8:13's Avatar
 
Join Date: Sep 2006
Posts: 1,250
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
I changed the mflowfps version in the first post, here it is. It's for maximum quality, not speed.

code removed
__________

Important note:
In the codes above I included this code: convertfps(29.976)
What convertfps does is correct what ffdshow does wrong.
What ffdshow does wrong is it plays the video back at the wrong speed, and this makes the video jerky every few frames.

What convertfps does is it makes the video run at the real fps speed the video is. So if your in Pal or Europe the changefps value would be 25: convertfps(25), if you have a mkv at 23.976 fps the you would set it to convertfps(23.976).
All NTSC DVD's I own I play from the disk and they are 29.976 fps, so the default value of convertfps in my codes is to what I would use when I watch my NTSC dvd's from the disk.

You can test this by using a frame doubler code and playing your video and with NO convertfps code, then pause the video and framestep the video. Do this for about 10 frames, the video will jump forward every few frames.
Then try with convertfps set to the framerate of the video and when you framestep it will be smooth with no jerky frames, like dominos tumbling.

And replying to mr 007, it's not just a mvtools with my visual studio settings. I removed code from it so it's not bloated with those functions.

I will list the functions I removed from the dll:

MCompensate
MMask
MSCDetection
MShow
MDepan
MFlowInter
MFlowBlur
MDeGrain1, MDeGrain2 and MDegrain3

The mvtools2.dll I made is in the ffdshow sticky thread. Along with a avisynth.exe I built where you just run the exe and your set. And there's a new mt.dll I made.

And to Subjunk. You missed the 2 codes I was proud of; test the one above.


There is new, and then you are new.
This is a moral of the bears and their cereal.
8:13 is offline  
post #1187 of 1968 Old 03-24-2011, 03:14 AM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by 8:13 View Post

In the codes above I included this code: changefps(29.976)

The correct framerate is 29.97, not 29.976.
23.976 * 1.25 = 29.97 and 59.94 / 2 = 29.97, it's all divisible.

Is there a reason you're using ChangeFPS instead of AssumeFPS?

Quote:
Originally Posted by 8:13 View Post

And to Subjunk. You missed the 2 codes I was proud of; test the one above.

It's getting better as it keeps getting closer to the one we already have, but still not there
SubJunk is offline  
post #1188 of 1968 Old 03-24-2011, 03:20 AM
AVS Special Member
 
8:13's Avatar
 
Join Date: Sep 2006
Posts: 1,250
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by SubJunk View Post

The correct framerate is 29.97, not 29.976.
23.976 * 1.25 = 29.97 and 59.94 / 2 = 29.97, it's all divisible.

Is there a reason you're using ChangeFPS instead of AssumeFPS?

It's getting better as it keeps getting closer to the one we already have, but still not there

I tried with the video at default or no changefps, info() showed the framerate to be 29.97. Then I did the framestep test I wrote about and it was jumpy. Setting changefps to 29.976 made the framestepping show a smooth flowing video; otherwise at 29.97 the character teleports when walking from one footstep to three footsteps where he should be, in one frame.

assumefps did not fix the problem when I set it to 29.976, changefps did. I should say I tested on the wolf clip I showed before, the clip that's 23.976 fps, it was being read as 23.97 before using changefps. I think oo7 wrote about that when watching Lady Gaga's video.


There is new, and then you are new.
This is a moral of the bears and their cereal.
8:13 is offline  
post #1189 of 1968 Old 03-24-2011, 05:03 AM
Member
 
travolter's Avatar
 
Join Date: May 2009
Posts: 162
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
about assumefps and changefps

- assumefps (change speed of video)
- changefps (duplicate every 4th frame)

for my taste the motion is much better with assumefps than changefps. The repeated frame of changefps stops the fluidity.

About Assumefps, it works pretty well when you need to slowdown videos.. I mean 30 to 25 .. 30 to 24.. 25 to 24.. but is not pretty good to speed up. Only works for small speed ups.. like 24 to 25.

Solution: Use 75hz in your monitor.. use assumefps(25000,1001) and interpolate with mvtools to 75000,1001

all my video sources are 24 25 or 30 and using assumefps(25000,1001) the conversions are done perfectly

edit.- If anyone have a working and low cpu usage way to convert 24 25 30 sources to 30.. plz let me know. I know that 75hz is not supported by many monitors and 60 is the default.

edit2.- Everyone have different eyes so its better try assume/change and use the more eye catchy for you
travolter is offline  
post #1190 of 1968 Old 03-24-2011, 04:01 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
I tested 8:13's MVTools2 build and found it to be slower than the official one.
I encoded 20 clips, with 10 of them using the official build and the other 10 using yours.
The average framerate for the official build was 18.66 while for 8:13's it was 18.17 which means 8:13's build performed 2.6% slower than the official one.

Quote:
Originally Posted by 8:13 View Post
I tried with the video at default or no changefps, info() showed the framerate to be 29.97. Then I did the framestep test I wrote about and it was jumpy. Setting changefps to 29.976 made the framestepping show a smooth flowing video; otherwise at 29.97 the character teleports when walking from one footstep to three footsteps where he should be, in one frame.

assumefps did not fix the problem when I set it to 29.976, changefps did. I should say I tested on the wolf clip I showed before, the clip that's 23.976 fps, it was being read as 23.97 before using changefps. I think oo7 wrote about that when watching Lady Gaga's video.
If you're right about your observations that means either the version of ffdshow you are using has a bug or there's something weird about the video/s you tested.

29.976 is not a real framerate
SubJunk is offline  
post #1191 of 1968 Old 03-24-2011, 05:13 PM
AVS Special Member
 
8:13's Avatar
 
Join Date: Sep 2006
Posts: 1,250
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Travolter, I learned from your script and applied lambda and lsad.

Here's my tweaked script:

Code:
code removed
Subjunk, could you compare this script against you Fast script, not the quality one.

I think Travolter, that if you do the test I wrote about and used your assumefps vs convertfps, you'll see a difference in results.

I would like to thank petitdragon again for his movie clip. It sure helps in tweaking the settings.

Efdit, on the movie Terminator 3 the fps doesn't need to be changed to 29.976, it works at 29.97, but on The Fifth Element and Ghost in the Shell Innocence the framerate needs to be 29.976 to pass the framestep test I mentioned.


There is new, and then you are new.
This is a moral of the bears and their cereal.
8:13 is offline  
post #1192 of 1968 Old 03-24-2011, 07:00 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by 8:13 View Post

...The Fifth Element and Ghost in the Shell Innocence the framerate needs to be 29.976 to pass the framestep test I mentioned.

Probably the reason you're encountering weird framerate things is because the files haven't been corrected for pulldown.
I just checked my official copy of The Fifth Element and it is definitely 23.976fps (as are 99.9% of movies, it is usually just things like sports that are 29.97fps).
If your copy is at a framerate other than that it means the pulldown hasn't been reversed, to fix that you can use tfm/tdecimate at the start of the script.
That will give you the proper framerate of 23.976.
SubJunk is offline  
post #1193 of 1968 Old 03-24-2011, 08:08 PM
AVS Special Member
 
8:13's Avatar
 
Join Date: Sep 2006
Posts: 1,250
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
How's that test of my script I just posted vs your fast profile? How does the code I just posted look, Travolter?


There is new, and then you are new.
This is a moral of the bears and their cereal.
8:13 is offline  
post #1194 of 1968 Old 03-25-2011, 08:15 AM
Member
 
travolter's Avatar
 
Join Date: May 2009
Posts: 162
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by 8:13 View Post

How's that test of my script I just posted vs your fast profile? How does the code I just posted look, Travolter?

I fixed your script because was not moving so smooth as mine..

so try with this now:

Quote:


setmtmode(2, 2)
video=ffdshow_source().changefps(23.976)
A = video
setmtmode(2)
super = A.MSuper(pel=2, hpad=2, vpad=2, rfilter=1, levels=0)
backward_vec = MAnalyse(super, isb=true, blksize=32, levels=0, search=3, searchparam=2, sadx264=6, pnew=50, lambda=1000, lsad=1200,global=true)
forward_vec = MAnalyse(super, isb=false, blksize=32, levels=0, search=3, searchparam=2, sadx264=6, pnew=50, lambda=1000, lsad=1200,global=true)
backward_vec_1 = MRecalculate(super, backward_vec, blksize=8, search=3, searchparam=1,pnew=50,lambda=250)
forward_vec_1 = MRecalculate(super, forward_vec, blksize=8, search=3, searchparam=1,pnew=50,lambda=250)
A.MBlockFps(super, backward_vec_1, forward_vec_1, num=FramerateNumerator(A)*2, den=FramerateDenominator(A)*1,thscd1=400, mode=0, thres=100)
GetMTMode(false) > 0 ? distributor() : last



The script have same fluidity as mine now: I changed levels=0, pnew=50, I used corrected values for lambda, and using mode=0
The problem comes with blksize=32. Its not enough for 720 or 1080 in very fast scenes with lot of details (I was testing with avatar sample video when the blue critters runs. they dont move so fluid as in other scenes).
Thats the cause I use blksize=8 and 4 in my script.

Anyway try to optimize .. its moving better each time
travolter is offline  
post #1195 of 1968 Old 03-25-2011, 03:26 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by 8:13 View Post

How's that test of my script I just posted vs your fast profile? How does the code I just posted look, Travolter?

I'm not interested in testing things against my Fast Preset, only against the Medium one. You can test it yourself easily
I can see several things in your code still that I consider to be mistakes, but I won't comment unless asked.
SubJunk is offline  
post #1196 of 1968 Old 03-26-2011, 12:46 AM
Member
 
neograniceni's Avatar
 
Join Date: Jan 2008
Posts: 20
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
CoreAVC 2.5 is out (http://corecodec.com/products/coreavc). Just tested it and too bad DXVA is not working with ffdshow RAW
neograniceni is offline  
post #1197 of 1968 Old 03-26-2011, 03:42 AM
Member
 
travolter's Avatar
 
Join Date: May 2009
Posts: 162
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by SubJunk View Post

I'm not interested in testing things against my Fast Preset, only against the Medium one. You can test it yourself easily
I can see several things in your code still that I consider to be mistakes, but I won't comment unless asked.

Subjunk, this is my fav script.http://www.avsforum.com/avs-vb/showp...postcount=1161 (720p sources)
Could you tell me if you see any mistake or something that could be improved plz?. I dont want tests or comparisons with your scripts because my script is for max fluid movement, so here the eye decide.
Im user that prefer maintain fluidity and not sacrifice it for image quality.
travolter is offline  
post #1198 of 1968 Old 03-26-2011, 05:49 AM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
I don't know if I understand what the difference is between fluidity and image quality.
It seems to me that when you lower the accuracy of the script the fluidity also suffers.
Can you explain what you mean?
SubJunk is offline  
post #1199 of 1968 Old 03-26-2011, 09:40 AM
Member
 
travolter's Avatar
 
Join Date: May 2009
Posts: 162
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
sorry for my bad english Ill try to explain better.

Fluidity=fluid movement, soap opera, truemotion for the eye, "image moves smooth as water" .. etc
image quality=no artifacts, blocks, details in image, sharper edges etc etc...

I know that many scripts published here try to Interpolate frames taking care of "image quality", so you are capturing screenshots showing image artifacts etc.

Im looking for the feeling to your eye.. more fluid image. The test and comparisons here have to be done looking to the video and check what one moves closer to the real thing.

You will say that all scripts are creating 60frames for a 60hz monitor.. but if you check your script and my script you will notice that movement does not have same fluidity= feeling to the eye (Im not speaking about artifacts/blocks and other image interpolation errors)
travolter is offline  
post #1200 of 1968 Old 03-26-2011, 05:32 PM
Senior Member
 
SubJunk's Avatar
 
Join Date: May 2010
Posts: 327
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
I can't see a difference in smoothness, this is really interesting though. Let's test it
I've made 3 test clips, can you watch them and tell me which is the smoothest?
Here they are:
1 - 2 - 3
SubJunk is offline  
Reply Home Theater Computers

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off