A little question regarding AD, ErMaC's and Zarxrax guide

[vemalas]

I read all the stuff, most of it understood, things i didn't will come to practice i guess.
Still one thing wasn't spelled there (or i missed), it's regarding colorspaces.

Steps i made and where i got stuck:
My raw videos are in YV12 colorspace. I opened them with AvsP, did a bit pre-cleaning and saved. Now. I open them one by one and cut the parts i need, save them in job control for later compression. Of course i choose Fast Recompress. In "Compression" i choose Lagarith.

Theres the question i had.

Should i export in YV12 with lagarith or RGB? I got this question because i've read that my footage must be RGB when importing to Adobe Priemiere. It doesn't support YV12. And then my move would be that final project export again from AP with lossy codec. Which one? RGB? And then convert to YV12 for final compresion in vdub?

But changing colorspaces are bad, yes? Step by step guide, or just simple logics would help indeed.

Thanks in advance.

[Scintilla]

Should i export in YV12 with lagarith or RGB? I got this question because i've read that my footage must be RGB when importing to Adobe Priemiere. It doesn't support YV12.

Adobe Premiere does all its internal work in an RGB colorspace*, but it's perfectly capable of working with video files in YV12 (for example, it will take MPEG-1s and XviD/DivX-encoded AVIs if you try, even though you really shouldn't be doing that).
I'd say go for YV12. If your original source was YV12 and you did all your pre-filtering in YV12, then there's no reason to convert to RGB and take up loads more hard drive space. Plus, I doubt that would work with Fast Recompress mode anyway.

Disclaimer: I have never worked with Lagarith. I guess it's possible that Premiere might have a specific problem with YV12 Lagarith files, but I wouldn't know.

And then my move would be that final project export again from AP with lossy codec. Which one? RGB? And then convert to YV12 for final compresion in vdub?

No. Your final export from Premiere should definitely be lossless.
As for what colorspace, you're going to have to convert to YV12 eventually anyway, so I don't think it makes a difference unless you want to use some filters that don't work in that colorspace (like the Dust family, which require RGB or YUY2).


* Unless you have Premiere Pro and YUV 4:4:4 video, but who uses that? (YUY2 is only 4:2:2 if I remember correctly.)

[Zarxrax]

Yea, there's always going to need to be 1 conversion to rgb and then back to yv12 again.

But as scintilla said, your editing program can convert from yv12 to rgb, so theres no need for you to actually save your lagariths as rgb (which would take about twice as much disk space).
Furthermore, lagarith has an option called "always suggest rgb for output". What this option does, is when you are using your lagarith in a program such as premiere, which prefers RGB video, then lagarith will convert the file from yv12 to rgb automatically as it decodes.
Now, what's the difference between letting premiere convert, and letting lagarith convert? I really don't know. Lagarith's conversion might be faster, or better quality. So I usually let lagarith do the conversion. Premiere has never been known for doing any sort of video processing in the "best" way, so I leave it to lagarith.

[vemalas]

Thank you for information scintilla and zarxrax, great community here.

By the way when i wrote lossy that was a typo, of course lossless ^^

Thanks again.

[Qyot27]

If I remember correctly, Lagarith's colorspace conversion code was taken from AviSynth.

Also to note, for Lagarith to maintain the colorspace of the video you give it, then you want to choose RGB (Default) from the drop down menu. Despite the name, it will still output YV12 if you give it YV12. The other options are to force Lagarith to do such a conversion. If you give Lagarith RGB and want YV12, then you'd choose the YV12 option, but RGB (Default) will give you what you put in*.

*although I highly doubt it would output anything other than YUY2, RGB(A), or YV12, so giving it YUV 4:4:4 (or 10-, 12-, or 16-bits per channel for any of the above it does typically support instead of 8-bits per channel) probably would force a conversion. But if you're using 4:4:4, you'd probably be better off using FFV1, since that one actually supports it.

Normally I edit using RGB-mode HuffYUV. Mainly because of the very fast decoding speed, and because I'd prefer Premiere not do anything to the colorspace.



In theory, yes, colorspace conversions are inherently lossy (although from how I understand it, this is sort of debatable when talking about scRGB and xvYCC conversions). However, two conversions shouldn't be visible under any sort of normal circumstances. Heck, I'd wager that even something on the order of 8 or 9 conversions would still be rather difficult to detect based on what kind of equipment you're targeting for final distro (TV, iPod, computer monitor, etc.). The loss incurred in a colorspace conversion is not anywhere near as noticeable as the artifacting produced by formats such as the MPEG-1/2/4 Part 2 (ASP, i.e. Xvid)/4 Part 10 (H.264/AVC), WMV9/VC1, Theora, etc.