Lossless H.264 isn't really lossless though, it forces chroma subsampling, and it still is only about as lossless as a full quality JPEG.
And the only codec I know of that doesn't force chroma subsampling is Theora 1.1 (Thusnelda) which isn't supported by a majority of players.
Build a man a fire, warm him for a day,
Set a man on fire, warm him for the rest of his life.
Joined: 11/11/2006
Posts: 1235
Location: United Kingdom
I was under the impression that lossless h264 had some problems under Quicktime on Mac OSX.
Also, runs don't need to use lossless h264. just high settings on x264. It doesn't need to be lossless to LOOK lossless.
<adelikat> I am annoyed at my irc statements ending up in forums & sigs
Is downloading higher quality movies at a larger filesize desireable?
Yes, I prefer that.
@ others: No, lossless does not provide a percievable quality gain over a high quality level like CRF20 in x264.
Go for CRF20 IMO.
OmnipotentEntity wrote:
Lossless H.264 isn't really lossless though, it forces chroma subsampling, and it still is only about as lossless as a full quality JPEG.
This is incorrect. Lossless H.264 encoding is lossless, full quality JPEG is not. The chroma subsampling you are referring to is caused by the colorspace conversion to YV12, which is not a part of the H.264 compression, but done prior to encoding. There is no way around it except use a different colorspace.
This is incorrect. Lossless H.264 encoding is lossless, full quality JPEG is not. The chroma subsampling you are referring to is caused by the colorspace conversion to YV12, which is not a part of the H.264 compression, but done prior to encoding. There is no way around it except use a different colorspace.
This is incorrect. Lossless H.264 encoding is lossless, full quality JPEG is not. The chroma subsampling you are referring to is caused by the colorspace conversion to YV12, which is not a part of the H.264 compression, but done prior to encoding. There is no way around it except use a different colorspace.
Or 4x the res at the cost of blockiness.
Yes, that can work around the apparent color information loss introduced by the conversion to YV12, but the conversion to YV12 is inherently not lossless :)
Yes, that can work around the apparent color information loss introduced by the conversion to YV12, but the conversion to YV12 is inherently not lossless :)
Does RGB -> YV12 conversion really matter, especially with the games which would most benefit from lossless compression (ie. the ones which use a few colors to begin with, ie. NES, GB...)? Even if it does change some color by one step, I assume the change won't be visible in practice.
Does RGB -> YV12 conversion really matter, especially with the games which would most benefit from lossless compression (ie. the ones which use a few colors to begin with, ie. NES, GB...)? Even if it does change some color by one step, I assume the change won't be visible in practice.
It matters a ton because of the small resolution. There are a ton of 1 pixel width or height things that when the color is loss, (1 color info per 2 x 2 pixel box) it really looks completely different. Maybe when I have time I can show a comparison.
You mean it doesn't simply take one RGB pixel from the input, perform a colorspace conversion to it, and then use it as a single pixel in the output, but also surrounding pixels are affected by the conversion?
Where can I find more info on that colorspace? I can't find anything named "yv12" at wikipedia. The closest things are YUV 4:2:2 and YUV 4:4:4, but I don't know if it's the same thing.
Ah, now it starts to ring a bell. It's similar to how JPEG is encoded by default, isn't it? In other words, the original image is converted to luminance (the brightness of the pixel) and chrominance (its color information), and then the chrominance channel is scaled down to half resolution on each axis (so that this channel has only 1/4 of the original pixels).
With regular photographs this loss of color information doesn't matter much because of how the eye works (luminance is much more significant than color information), but it might indeed have a large impact on pixel graphics (especially when the overall amount of colors is small).
It's odd, though. IIRC the JPEG format supports not downscaling the chrominance channel resulting in better color preservation (at the cost of a slightly larger file). I wonder why x.264 doesn't support that too (being a much newer format and all).
If you ask me, I think that torrents should have a higher quality*.
But I also think that it is worth considering adding multiple files to torrents. I do not think that it is generally something that should be done. But there are exceptions. For example, runs that have multiple encodings. Like pure vs camhack and so on. Things like author voice comments should also be included if possible.
One other thing that could be included that struck my mind is the actual movie file, but while I think that the extra size would be trivial, the actual use for it is slim. The people who are capable of using it will most likely not even want the encoded version. And the rest will just find it to be clutter. So yeah, I do not think that it should be included.
*I am not qualified to argue about what quality that is.
