--tune grain on BluRay sources

[nibus]

I've noticed that most BluRay movies are mastered with a lot of grain. I've always used --tune film on my encodes but am wondering if the grain setting may help with BluRays and prevent banding/blocking.

Does anyone here use mostly --tune grain and if so, does it usually require a higher bitrate, and do fast action scenes suffer due to bandwidth allocation when compared to --tune film?

[nibus]

Okay so I've run some tests on --tune grain vs --tune film with some BluRay sources.

Tune grain definitely helps keep a little more detail, and more grain (imagine that?) As a result, a higher bitrate is required, and with CRF so far it has required about 25% more bitrate to keep the same quality.

High motion scenes look fine so far. So I guess it comes out to whether you want all the details and bits of grain preserved along with slightly larger file sizes.

Here is a comparison of film vs grain screenshots:

film:
http://dl.dropbox.com/u/5637223/IronMan%20old.mkv_snapshot_00.02.32_%5B2010.07.19_22.33.58%5D.png

grain:
http://dl.dropbox.com/u/5637223/IronMan.mkv_snapshot_02.32_%5B2010.07.19_22.33.26%5D.png

It's a small difference, but when expanded to 1080p it becomes a little more apparent.

[dansrfe]

yea I've always wondered whether to use --tune grain or --tune film on BluRay sources. What's a good crf value to use on BluRay sources anyway?

[obieobieobie]

Look at the eye lid of the soldier. That's where I see the biggest difference.

[Dark Shikari]

Comparing quality at different bitrates is pointless.

[nibus]

yea I've always wondered whether to use --tune grain or --tune film on BluRay sources. What's a good crf value to use on BluRay sources anyway?

I use CRF 16, but that's only after a 3 way denoising chain -


TTempsmooth(maxr=7, fp=true, strength=1, lthresh=3, cthresh=4)
SMDegrain(tr=1, thSAD=100,blksize=8,overlap=4)
dfttest(dither=3,sigma=3,sigma2=3,tbsize=3)
LSFMod(defaults="slow",strength=50)

If you were to go that low without any denoising the file would be quite large. With these light denoisers a 2 hour movie usually lands right around 4000kbps at 720p, rounding out to about 4gb total.

Comparing quality at different bitrates is pointless.

Well part of the point was not necessarily quality but how much more bitrate --grain takes in comparison. I will run a 2 pass at a specific bitrate with both tunings and post that as well, because obviously the --tune grain image is using more bitrate (about 500kbps).

[laserfan]

Tune grain definitely helps keep a little more detail, and more grain (imagine that?) As a result, a higher bitrate is requiredI've done no tests, but was of the impression that using "grain" over "film" was the selection-of-choice for those movies that are designed with a noticeable amount of grain in them (examples "300", Saving Private Ryan, the TV series "Damages"). And perhaps for older movies where they were shot with a grainier grade of film stock.

So I use --tune film most of the time and don't otherwise think about it.

Ironman is a newer film with very fine grain so I wouldn't use --tune grain with it.

[nibus]

Well part of the reason I was considering using it is because I use Dfttest's dithering, which adds little bits of grain to avoid banding and blocking.

[Keiyakusha]

You can't save that DFFTest's dither. This dither is just not good enough. It's unbelievably small like 1px dots and gets killed even by resize so even if you'll manage to keep it - it can be lost on playback :Р

[nibus]

You can't save that DFFTest's dither. This dither is just not good enough. It's unbelievably small like 1px dots and gets killed even by resize so even if you'll manage to keep it - it can be lost on playback :Р

Huh... in my experience it's been extremely effective. But I've only used it on BD film sources, no animation.

[Keiyakusha]

Well of course if you using something like dither=3, it won't be killed entirely, but I can't imagine where such settings is not overkill since even on totally flat animation dither=2 is enough to fill all gradients. Normally you shouldn't use more than dither=1. For BD film sources I haven't found much use of this dither at all. Such films already contain enough amount of grain to look good so dfftest just replacing bigger grain (which is easier to compress and which is meant to be there) with smaller grain (which is harder to compress and which is not meant to be there).

[nibus]

My whole issue with BD grain is that it is too big and ugly. That's why i replace it with Dfttest's dither, which looks much better especially on 720p encodes. IMO / YMMV of course.