The H.264/AVC/MPEG4 discussion thread

[Shimapan]

Originally Posted by tyranuus

@Pantsu, I'm no expert so couldn't give you much feedback, however I noticed you opted for QP when I took a peek at the XML (as I don't use Avidemux)
How much difference QP makes over CRF I don't know,but the reason I opted for CRF is I'm constantly seeing it recommended all over the place as achieving better file sizes at equivalent picture quality.
Example (3rd post):
http://forum.doom9.org/archive/index.php/t-107687.html

You didn't look properly then, because I'm using Constant Rate Factor, not QP.
I'm generally using a Crf between 16-18.

[Desbreko]

Originally Posted by tyranuus

--preset slower --tune animation --level 4.1 --crf 16.5 --fps 24000/1001 --aq-strength 1.0 --psy-rd 0.8:0.1

--preset slower--level 4.1 --crf 16.5 --fps 24000/1001 --bframes 5 --psy-rd 0.8:0.1

The only difference between these two is that the first one will use --deblock 1:1 from --tune animation instead of the default of 0:0. It will be a pretty small difference unless you're encoding a grainy source, though.

[tyranuus]

Originally Posted by Shimapan

You didn't look properly then, because I'm using Constant Rate Factor, not QP.
I'm generally using a Crf between 16-18.

Apologies then, I was reading the XML directly, and one of the first things it says is basically x264 encoding mode AQP not CRF so you can understand where my misinterpretation came from.

@Debresko - cheers for the clarification

[zrdb]

And as the next new shiny Cadillac with god plated trim rolls into town everybody's gonna jump all over it.

[zrdb]

Originally Posted by Quarkboy

And what does DS think the prospects of seeing a cheap, consumer targeted 10-bit capable decoder chip are in the new future?

I can say with authority having conversed with Sigma directly that no current or near future soc based media player has or will have the ability to decode 10 bit video. They don't see any need for it as anime is a niche market-so I had no choice but to put my trusty PCH A210 out to pasture and replace it with a shitting htpc that I built myself. Ain't progress wonderful?

[tyranuus]

Hi Guys, me again

I'm still using the settings put together here previously last time I came around to ask a few questions, but, and this might potentially be a loaded question dependant on viewpoints, but how are you finding the shift to Hi10p is affecting your encodes?

With x264 Hi10p encoding support now starting to mature (I'd hope), are you guys just dropping in the 10bit encoder variant exe and using the settings you used before, or has it changed any of the recommendations that previously would be suggested for those using the 8bit encoder.

Alternatively, do you guys think it's still immature enough that those of us using it to go through old collections or encoding now and then, are best off sticking with the good ol' mature 8bit? From doing a little quick testing of my own, I've noticed a sample clip I produced 8bit@CRF 16.5 and 16, and Hi10p@CRF16, produced file sizes ~10% smaller than the 8bit equivalent, but probably encoded about 25% slower.

I appreciate any feedback!

[Dark Shikari]

The only typical changes are that you typically don't have to worry about banding with 10-bit -- so you can often drop any weird tweaking (psy-trellis, etc) that you may have had to try to better retain dither.

If your main quality issue was one of banding, you can also raise CRF a whole bunch and get away with it.

[Desbreko]

What DS said, pretty much. You can use lower psy-rd/psy-trellis for non-grainy anime with 10-bit since they aren't needed for retaining dithering as much. This is nice because it doesn't cause as many artifacts around the line art and uses fewer bits.

I still like to use a little psy-trellis, though, as I find it's nice for keeping small details sharp during motion. I've taken to using --psy-rd 0.6:0.05 for most anime.

Otherwise, business as usual and just enjoy the lower file sizes.

[tyranuus]

Thanks for the feedback, if it's not played any major hassle I'll probably just stick with the command line I was using previously and enjoy a 10% or whatever hit on the file size.

[ANIKIN]

Guys, where can I learn to encode on CMD?

[j0x]

Originally Posted by ANIKIN

Guys, where can I learn to encode on CMD?

here is a tutorial i made -> https://sites.google.com/site/taumox...n-command-line

im not expert in encoding by the way i just happen to share what i have learned from the people who knows better

[Nessaja]

I'll just hijack this thread for a few questions;

what's the general limit before up-scaling 720x480 content becomes rather useless?

For 16:9 I go with 848x480 and 704x480 @ 4:3. Or is there any gain from going say 1024x576 and 960x720? As far as I've noticed, blocking becomes worse not to mention pixelation. But I assume that's as far as you can get with filtering? Or is the bitrate to low?

[max2k]

Blocking should not be an issue of the upscaling itself, if its not in the source before you encode it, it should be the bitrate. If you upscale you need more information for the increas of pixels. Filters are ther, to remove artifacts that are in the source from the beginning and they don't help if the macroblocking comes from ure encode. Ok it could help to smooth the source plain. It should be easyer to compress, but grrr~...

But why upscale in the first place. If you not go all mad on Supersampling you should not get any gain. Even if you supersample like a god, you don't get more details.

What Codec do you use? For x264 you dont do a 2pass at target bitrate normaly you do a 1pass constant quality (crf) encode and only fall back to 2pass at target bitrate if the video stream needs too much bitrate.

Rant:
Still wonder how someone could encode Bleach in 720p and need 500mb for the Vid-Stream with no gain in quality. Must be, to get more Leechers then the other Group ...

[PositronCannon]

Seconding the "why upscale" part. Might as well just let the player do it in real time if you really want a bigger picture. Even if you sharpen it to death, the drawbacks from that mean the only thing you'll actually gain is a higher filesize.

[Desbreko]

About the only point of upscaling 720x480 is to avoid dealing with anamorphic video. You can use nnedi3_rpow2 if you upscale upon encoding instead of playback, but even that will only be marginally better unless you're upscaling extreme amounts.

[Nessaja]

Well.. I'm also playing with 320xXXX and 480xXXX sources, so it becomes rather useful to upscale and filter rather than having the player handle it. Not to mention it's not something that can be played OOTB.

[Quarkboy]

Originally Posted by Nessaja

Well.. I'm also playing with 320xXXX and 480xXXX sources, so it becomes rather useful to upscale and filter rather than having the player handle it. Not to mention it's not something that can be played OOTB.

Oooh, superscaled filtering is a whole different matter. There's a lot that can be said for a superscaled warpsharp on SD sources... (if done right, anyway)

[Desbreko]

Supersampled filtering (i.e. filtering at an upscaled resolution) is really useful, yeah, but you don't need to keep the final output clip at the supersampled resolution. A number of common filters actually use supersampling internally but then just scale the clip back to the original resolution for their output.

[Nessaja]

So when upscaling to say 720p and uploading to youtube? (source is 480x272). Oh and should I use Lancoz or Spline when upscaling content? I've used Lancoz3/4 for ages but perhaps I'm just doing it wrong? Oh and 1 or 2-pass?

[jfs]

Originally Posted by Nessaja

perhaps I'm just doing it wrong?

Yes. For this reason:

Originally Posted by Nessaja

upscaling