vfr to cfr convertion

[mathmax]

Hi

I would like to convert a vbr file to cfr and I wonder:
1. if the process only duplicate some existing frames or if it create new frames by interpolating the existing ones (which would give a smoother video)
2. if the conversion process needs a reencoding (i.e a compression with a loss of quality).

Also, they say here that it's either possible to either convert to cfr 120fps or 30fps. In case I choose 30fps, the video should be "re-decimated" but I don't know what it means exactly... nor I don't know which are the advantage of each method given in the wiki-tutorial.

Thank you in advance

[poisondeathray]

Hi

I would like to convert a vbr file to cfr and I wonder:
1. if the process only duplicate some existing frames or if it create new frames by interpolating the existing ones (which would give a smoother video)
2. if the conversion process needs a reencoding (i.e a compression with a loss of quality).

Also, they say here that it's either possible to either convert to cfr 120fps or 30fps. In case I choose 30fps, the video should be "re-decimated" but I don't know what it means exactly... nor I don't know which are the advantage of each method given in the wiki-tutorial.

I'm only familar with the convertfps method ; when inserts are required it inserts a blend (not a mvtools generated "in between" frame), and requires re-encoding

by "re-decimated" it means by using 30 fps instead of 24 fps, it will insert extra frames (when you use convertfps=true, it adds extra frames). So for film content (24fps) , you have to decimate 30fps back to 24fps . ie. you have to get rid of the dupes later

[mathmax]

Thank you

Why not convert directly to the final frame rate. For example, if I want 24fps, why would I first convert to 129fps and then decimate to 24fps? Why not convert directly to 24fps?

You said that the convertfps requires reenconding, but I guess the DirectShowSource() doesn't compress anything.. I mean there shouldn't be any loss of quality on the existing frames after executing the function, right?

[poisondeathray]

Why not convert directly to the final frame rate. For example, if I want 24fps, why would I first convert to 129fps and then decimate to 24fps? Why not convert directly to 24fps?

I don't know. I use the method directly to final FPS. I think it's for a common framerate , because hybrid sources will have 30fps (video) and 24fps (film), so lowest common multiple is 120

You said that the convertfps requires reenconding, but I guess the DirectShowSource() doesn't compress anything.. I mean there shouldn't be any loss of quality on the existing frames after executing the function, right?

using avisynth requires re-encoding; it frameserves uncompressed video . No quality loss but huge filesizes unless you use lossy compression

[mathmax]

ok, thank you.

But I have a problem with the DirectShowSource() method. When I load my video using this function, I get a grey screen in vritualdub. My file is a .mkv containing a h.264 file and a timecode v2.

In the avisynth tutorial it is said :

if the source is mkv, you can also use mkv2vfr and AviSource.

Maybe I should rather use these functions. But with AviSource I don't have FPS parameter... and mkv2vfr is an external program and I don't know if it generates a lossless AVI output file...

What do you think I should do?

[poisondeathray]

install haali media splitter, ffdshow, enable h.264/avc in the configuration. On win7, use preferred filter tweaker as well

[mathmax]

That's perfect Thank you, you helped me a lot.

Just one last question: having a set of images (for example .bmp) and a timecode file. Is it possible to wrap them into a vfr lossless .mkv? Which tool would you use for that?

[poisondeathray]

I don't think mkv supports bmp , so you could use ImageSource() to encode to some video format, then mkvmerge with timecodes

BMP's are RGB, most video formats are YV12. I don't know if MKV supports lossless RGB video formats (it accepts uncompressed RGB)

[mathmax]

Thank you. Will I loose quality during the the RGB to YV12 conversion?

[poisondeathray]

Will I loose quality during the the RGB to YV12 conversion?

Yes. Chroma is subsampled (lower resolution color)

If your original was YV12 flv, then skip the BMP stage. YV12<=>RGB conversions are lossy

[mathmax]

mmm.. are there other formats of image lossless and YV12?

[poisondeathray]

No. Most lossless image formats are RGB (png, tiff, exr, etc...)


There are plenty of lossless YV12 video formats (lagarith, huffyuv, ut video codec etc...) , and plenty of lossless RGB video formats (same, lagarith etc....). Some of the lossless codecs have different modes (YUY2, RGB, YV12 etc...)


The point is to stay the same format (avoid changing the original format). If original is in YV12, stay YV12. But sometimes it's unavoidable quality loss (e.g. you need to decompress frames for edit in photoshop or after effects for example - these work in RGB (or CMYK) - so there is no choice)

Chroma subsampling is not a huge loss, most of the time you will barely notice it unless you look closely , or certain types of content like video games, animation, where color borders are distinct (human eye is less sensitive to color than luminance)