Trick to keep film grain w/o high bitrate?

[Episodio1]

Hi! Is there any filter or config in XviD to keep film grain at good qual without using insane high bitrate? :)

Thanks.

[BigDid]

Hi,

Usually not my concern, I end with little sharpening if too smooth;

Some links after making a search on addgrain:
http://forum.doom9.org/showthread.php?t=97217&highlight=addgrain
http://forum.doom9.org/showthread.php?t=84181&highlight=addgrain
http://forum.doom9.org/showthread.php?t=80961&highlight=addgrain

Make more search on it (grain ,noise, sharpen, compressibility etc...), good luck.

Did

[Caroliano]

The best solution is put the grain back with post-processing, with the noise option in FFDShow for example. You can try to use a high-bitrate matrix like 6of9 and see what hapens. It shoud keep more grain than h.263. But as the definition says, it is an high-bitrate matrix, I don't know what hapens to your movie using it in an noise source with low bitrate. You may have to use an real high bitrate because gain is random, so, a big task for the encoder.

BigDid: I assume that he want's to keep the grain, so he already has it in the source.

[IvS]

The request was how to keep the original video's grain using as low bitrate as possible.
Grain can cause a huge rise in bitrate. Often, when the bitrate is restricted to less than huge, the resulting video's quality is very poor. "Retrieving" that grain with post processing is definitely not possible, since the video that has a too low bitrate is smoothed out poorly. Suggesting a smoothing filter is unrelated as well, since many videos have parts where the grain plays a major role, as in, it is an intentional and integral part of the video. And sometimes, of course, it's not the entire screen that has the grain.

Some people may have the idea that grain only belongs to old films. That's definitely not true, many people use various types of grain today increasingly to bring a distinct feeling to a video. Grain has become a real trend in 3D demos lately, and it can look really great.

Here (http://www.pouet.net/prod.php?which=20209)'s a fine example of one where grain plays major part. I captured it at 60fps using .kkapture (http://www.farb-rausch.de/~fg/kkapture/) using Lagarith and tried converting it to XviD, but XviD nor DivX could handle it at all with any settings. No matter what, the bitrate handling is too poor due to the harsh dominant grain.

So in addition to the first question, I'd like to know not how to not get a huge bitrate, since I suppose there's no other way sometimes, but, how to get a not too low bitrate with content like the above :).

[Didée]

... I'd like to know not how to not get a huge bitrate, since I suppose there's no other way sometimes, but, how to get a not too low bitrate with content like the above :).
Try that:

Take 6of9 (http://home.arcor.de/dhanselmann/_stuff/SixOfNine.xcm) or 6of9-HVS (http://home.arcor.de/dhanselmann/_stuff/SixOfNine-HVS.xcm) as quantization matrix. Constant Quant=2. Bframes either none, or max.1 @ 1.00/1.00. VHQ=2, B-VHQ yes, Trellis yes, Qpel yes, Chroma motion yes. If grain is strong: ME=4 (yes indeed, "four"!)

Should solve your problem of "not achieving enough bitrate", and deliver a close-to-perfect result.

(I know the combination ME=4 + VHQ=2 seems idiotic, but I've made good experiences with that in some very grainy cases ... of course this draws bitrate, but that's what you want, no?)

[Sharktooth]

uhm, i'd try EQM V3 UHR without messing with impossible settings and if that's not enaugh to achieve the wanted bitrate i'd try EQM V3 EHR.

[IvS]

Thanks Didée! That matrix indeed makes the bitrate handling much better for this kind of content.
ME=4 is not stupid at all, I always use the most exhaustive settings possible :).
Constant quant 2 seems unneeded for it though, I'll try 2 passes and aim for something around 3 and for b frames 4.

Sharktooth, thanks, I'll try that as well.

[IvS]

OK, so I see it's not related to the matrix as much as it's related to the fact I used 2 passes. With either matrix (or without), when using 2 passes, the resulting quantization values are way too high. Sharktooth's matrices and 6 of 9 do seem to provide a somewhat more sensible bitrate handling with a constant value. (I've studied this video's parts thorougly so I know more or less how requiring each part is :))

Edit: Surprisingly (or not so surprisingly since h.263 often performs better with high motion content), now that I tried h.263 with the same constant value (3.5) it seems to provide the same quality at about half the bitrate. I assumed it would blur the grain more but it doesn't.

[Didée]

uhm, i'd try EQM V3 UHR without messing with impossible settings
It's not "messing", and the settings are not "impossible".

A silken cloth deals badly for cleaning dirt ... and with "everything @ max.", XviD's ME is oversophisticated for strong-grained sources.
Less vector bits that would only make textures crawl, but more texture bits to reproduce what can't be compensated anyways. That's the grainy way.

BTW, we can put all EQMxxx's and 6of9's in the trashbin, anyway. Just today it was revealed that H.263 gives better quality at half the bitrate ... d'oh! Having known this, I could've made all those anamorphic full-size encodings at 600 kbps! <snort>