YV12 FAQ

[HarryM]

Originally posted by sh0dan
When I open MJPEG files with YV12, using Morgan, it has swapped the UV channels (giving me I420 instead of YV12). I'll do a SwapUV function shortly. I'm using an older version of Morgan, so it might be fixed in the latest version.

Thanks for answer.
Can you implement 'SelectRangeEvery' function (for fast compresibility check) too?

[sh0dan]

How is it different from selectevery?

[vlad59]

SelectRangeEvery was included in mpeg2dec.dll by Dividee. I think it's now removed from mpeg2dec3 (by MarcFD) so adding this filter in YV12 is juste a matter of compile.

I can do that tonight.

[sh0dan]

Couldn't it simply be included in AviSynth 2.5 (or is that what you meant). The old filters no longer works anyway, so why not just add it as an internal feature?

[HarryM]

Originally posted by sh0dan
Couldn't it simply be included in AviSynth 2.5 (or is that what you meant). The old filters no longer works anyway, so why not just add it as an internal feature?

SelectRangeEvery is 'must be' for GordianKnot (compresibility check in this famous rip program). I know, actual version GK isn't compatible with Avisynth 2.5, but future...

[HarryM]

Originally posted by HarryM
SelectRangeEvery is 'must be' for GordianKnot (compresibility check in this famous rip program). I know, actual version GK isn't compatible with Avisynth 2.5, but future...

I am not (big) programmer, but I think, that implementation of 'SelectRangeEvery' direct into Avisynth2.5 is very easy (few lines of code).

[vlad59]

Yes it should be included in avisynth 2.5 .
If you need the original code I can send it to you (adding it to field.cpp is just a matter of copy/paste) or if you don't have time I can do this for you.

[HarryM]

Originally posted by sh0dan
How is it different from selectevery?

SelectRangeEvery is very similar to SelectEvery.
SelectEvery works with alone frames, SelectRangeEvery work with intervals of frames.

E.g.

SelectRangeEvery(280,14)

Select this frames
0 to 13, 280 to 293, 560 to 573, etc.

[HarryM]

Originally posted by vlad59
Yes it should be included in avisynth 2.5 .
If you need the original code I can send it to you (adding it to field.cpp is just a matter of copy/paste) or if you don't have time I can do this for you.

Sorry, but i have no experiences with C++ programing and with team programing.

I use relative often VisualBasic or sometime Delphi only.

[vlad59]

@HarryM

Sorry while posting I was answering to Sh0dan but in the meanwhile you posted two answer too.

So I'll make the modification. don't worry.

[sh0dan]

@vlad: I'm doing modifications on field.cpp (I think there is a bug in Separatefields()) today, so feel free to mail or PM me the sources.

[JohnMK]

If someone could make available the SelectRangeEvery-capable binary of mpeg2dec3.dll, I'd be so appreciative.

[vlad59]

Originally posted by JohnMK
If someone could make available the SelectRangeEvery-capable binary of mpeg2dec3.dll, I'd be so appreciative.

SelectRangeEvery will be added in avisynth 2.5's core. So we just have to wait till a new alpha release of avisynth dll.

[JohnMK]

Could that be today please? This is so fundamental.

[vlad59]

I checked CVS and this function is already added. I won't be able to compile before 5 or 6 hours.

And I don't know if it a really good idea to compile this fresh checkout (I don't know if Sh0dan is modifying something).

[JohnMK]

How about the old compile + this function?

[SansGrip]

Question 6:

FluxSmooth supports 2.5/YV12 as of 0.2 .

[Richard Berg]

Be very careful - you cannot use it after a crop, since ffdshow assumes that input pitch = output pitch! This will give the wellknown distorted pictures! I hope someone else has the time to take up the challenge?

Is it more complicated than replacing every instance of 'width' with 'pitch'?

Anyway, I suppose it's about time I figured out how to work with the CVS branch...

[JohnMK]

I just wanted to thank everyone who's made this possible. My encodes are easily -- easily 20-30% faster, just as you say. And quality doesn't look any less. In fact, theoretically, it's better, right?

[sh0dan]

Originally posted by Richard Berg
Is it more complicated than replacing every instance of 'width' with 'pitch'?

Anyway, I suppose it's about time I figured out how to work with the CVS branch...

1) Unfortunately, yes. It uses both pitch and width, but it only uses ONE pitch. And when input and output images have different pitch, the pictures it distorted. I have an "almost" working version, but there are still one wrong pitch I'm having troubles finding.

2) It's really easy, just use the "avisynth_2_1" branch name to get access to the files.