Avisynth - What's Wrong with this Script. (FPS Convert/Motion Estimation)

[RAB78]

I'm now running 3 instances of Virtualdub, each running a batch job of 8 files. Should get through these files in no time.
I might try running 4 at a time and just keep and eye on my CPU temp. Very glad i can do this. Gonna save a lot of time.

In regards to pel....
I'm doing all my videos at pel=1.
If i notice a scene with major distortion in it, due to the motion compensation, will changing pel improve the overall conversion quality?
At the end of the day, does it really give better results and is it really worth the extra encoding time?
Thanks a lot!

[poisondeathray]

pel=1 is faster (full pixel) . pel=4 is slower (quarter pixel)

Sometimes one is noticably better than the other (in terms of artifacts). You probably don't have time to adjust settings per scene, it sounds to me like you want to set and forget

If you have other computers in the house (or even friends' houses), consider "borging" them to the cause temporarily. Parallel processing is the fastest and safest way to speed up avisynth processing, not MT

Critical scenes might require special attention (maybe something very sentimental or important), that's where you might consider adjusting settings or resorting to other methods

[newpball]

If i notice a scene with major distortion in it, due to the motion compensation, will changing pel improve the overall conversion quality?
At the end of the day, does it really give better results and is it really worth the extra encoding time?
Thanks a lot!

Don't know the quality of the avisynth script I use Twixtor.

However what is important is to recognize scene changes. I do not know if the Avisynth script handles that. On the other hand if the video is a raw selection of cuts it probably does not matter assuming the cuts are with enough lead time.

[jagabo]

In regards to pel....
I'm doing all my videos at pel=1.
If i notice a scene with major distortion in it, due to the motion compensation, will changing pel improve the overall conversion quality?

When you see big distortions it's because motions were misidentified. Using sub-pixel accuracy usually doesn't help that. It's for getting smoother motion at small scales.

[poisondeathray]

The "best" avisynth optical flow plugin IMO, is SVPFlow. IT uses a modified mvtools2. It tends to give slightly cleaner results, less edge morphing. It's not going to be a huge difference in terms of quality, but slightly better IMO. But one big plus is it can use GPU acceleration (might be roughly 10-30% faster)

http://www.svp-team.com/wiki/Plugins:_SVPflow



Twixtor, when used at default settings tends have the same problems as mvtools2. All optical flow approaches do. What separates twixtor (or at least the pro version) , is the ability to guide the motion estimation with mattes, trackpoints, splines. This ability for user intervention is what makes it probably the "best" IMO.

[RAB78]

Thanks for the info guys.
I've come across SVPFlow while searching for info online about motion interpolation. It sounds great but i wouldn't know where to start with it.
I've come so far since i first had this problem and i now have a script that works well for me.
I'm not sure i could start over again unless someone wants to put together a script for me to try?
I think i have all the necessary files in place. It's up to you.

[RAB78]

Hi again, I'm still converting my files. Lol.

I'm trying to convert all my 25p footage to interlaced but the result appears to be progressive still.
I can see none of the obvious combing artefacts usually present in interlaced footage and MediaInfo says it's progressive.
However, Vegas MS says it's interlaced.
I just wanted to check that i'm doing the script right for interlacing 25p.
Here's my script:

DirectShowSource("J:\Scotland 2014\Canon 550D\02-08-2014\MVI_3702.MOV", fps=25)
Changefps(50)
AssumeTFF()
separatefields()
selectevery(4,0,3)
weave()

Thanks!

[poisondeathray]

For SVPFlow, you just copy & paste the script from the instructions on the page. If you get an error message , post the message. I wouldn't worry about it too much if you can't get it working, it's not *that* much different - you will still get occlusion artifacts, edge morphing etc...

Hi again, I'm still converting my files. Lol.

I'm trying to convert all my 25p footage to interlaced but the result appears to be progressive still.
I can see none of the obvious combing artefacts usually present in interlaced footage and MediaInfo says it's progressive.
However, Vegas MS says it's interlaced.
I just wanted to check that i'm doing the script right for interlacing 25p.
Here's my script:

DirectShowSource("J:\Scotland 2014\Canon 550D\02-08-2014\MVI_3702.MOV", fps=25)
Changefps(50)
AssumeTFF()
separatefields()
selectevery(4,0,3)
weave()

Thanks!

source fps=25 , and changefps(50) will duplicate frames (every 2nd frame) . ChangeFPS() duplicates or deletes frames.

So the content will be 25p duplicated. When you re-interlace that, the content will still be progressive because each field represents the same moment in time. ie. No new temporal information had been generated . Compare that to interlaced (true interlaced [i] content[i] - where each field represents a different moment in time. That is the reason why you see "combing" in the first place on interlaced content. You need 50 moments in time per second to generate 25frames per sec interlaced - you don't have that with changefps()

[newpball]

I'm trying to convert all my 25p footage to interlaced but the result appears to be progressive still.
I can see none of the obvious combing artefacts usually present in interlaced footage ....

Why would you?

The only time interlaced footage would show artifacts is when the two fields forming a frame represent a different point in time. For a progressive source both fields obviously reference the same point in time.

[RAB78]

@poisondeathray
I'll try the script on the page and see how i get on.
It'll be good to get it working if it'll be a bit faster than the normal MVTools.
So your saying that i should use the motion estimation method instead of ChangeFPS()? That's a shame. I was hoping to convert the 25p files quicker.
.........
I've just tried the motion estimation method and now i can see that it is interlaced. Is this the right way to convert 25p to 25i? There's no other way?
Thanks for your very clear explanation of what is occurring in my script.

@newpball
This is why i'm asking for help.
To be honest, I didn't know i had to create two fields a frame. I think i misunderstood the concept of what i was trying to achieve.
I was getting 12.5fps interlaced first time i tried it so i thought that if i doubled the framerate, it would sort it out.

[jagabo]

I was getting 12.5fps interlaced first time i tried it so i thought that if i doubled the framerate, it would sort it out.

That would have worked -- except the method you used to double the frame rate simply duplicated each frame. So when you pulled fields down from those frames you were back where you started.

[RAB78]

Thanks Jagabo. Good to know i was on the right track.