Regarding hi10 Encodes

This is to the hi10 encoders:
All right, first off, I'd like to say that I've been a fan of hi10 site for a long time. I have downloaded lots of stuff and feel sad that so many issues are cropping up now. I was intrigued by your low sized good quality encodes and I learnt encoding myself, and have been doing it for quite some time now. What I have to say is regarding encoding itself.
I'd prefer if you take it as constructive criticism, though I don't claim to be an expert.
Well, there is nothing to complain about the quality of encodes, at most places. Please do not get me wrong: I don't mean to undermine anyone but I'd like to know why you(the encoders) do not change any encoding parameters. I may be wrong, but in most of the encodes I downloaded from your site I noticed the same parameters: crf 24(or lower) psy 0.4:0.0 ref 16 b-frames 10 deblock 1:2 merange 24 aq 1:0.6 . The thing is that these parameters work fine for most stuff but not for all. For instance, these same settings were used for your BTOOOM encodes, and the flat dark regions did not come out well at all. I encoded the same episode with different settings and I got a lower size with a good quality. On the other hand, there were things like Ano Hana where the file sizes were too high, unnecessarily. Kimi ni Todoke was encoded at 23 crf I think, which I felt was not necessary. So a few questions:
I know that all the encoders here are good and experienced and know their job well. So I'd like to know why they use the same parameters for everything that I've seen.
1. Why not use a higher psy than 0.4 and different aq (say 1.0, mode 3) for darkish anime?
2. Why are there no encodes above crf 24? I thought the point was to give high quality, low size anime; so if you are getting the same quality with crf 25, (which is the case with a decent source mostly), then why not?
3. I may be wrong here but why don't I see any filtering? Things like gradfun are needed at some places and the encodes do not look good without it. I've done some encodes that looked crap at dark places even at crf 19, without adding grain.
4. Why no zones anywhere? That way you can use higher crf for the rest of the episode and get lower sizes.

Don't get me wrong: I know you have tight schedules and you know what you are doing, but I just thought I'd ask you what has been bothering me for quite a while. I do not doubt your quality but merely want to know why. I'm asking this merely because I've found you people very open to what others have to say and I like this site a lot. I didn't know where else to post so forgive me if I got the wrong place.

A month ago another user came up with the same topic, I'm going to show you the same comparison that I made back then:

Sharpened encode (SeMi) vs Deblocked encode (Hi10):
http://screenshotcomparison.com/comparison.php?id=31903

Source vs re-encoded:
http://screenshotcomparison.com/comparison.php?id=31907
http://screenshotcomparison.com/comparison.php?id=31908

That's it, you know what's the difference? 20% file size.

Hi, thanks for replying. As I said, I didn't know where to post because I thought it was a general observation for most anime. I was talking about the anime I downloaded over the last year(saw in SAO, Steins;Gate, BTOOOM, Kimi ni Todoke, Danshi Koukousei, Kanon, FLCL, Gintama, Sakurasou- all have exact same parameters including crf 24.  There are more but I'm too lazy to use mediainfo again. I'm doing crf 25.5 to 26 depending on episode for Sakurasou and for SAO BDs and the encodes look flawless to me with much lower sizes). Out of joblessness, I was comparing how hi10 had encoded the ones which I myself did and maybe hastily jumped to the above conclusion. Sorry if I was mistaken. I was shocked because I even vary all parameters from episode to episode of each anime and I saw the same thing for all anime here- be it kanon or btooom.

Hmm personally I wouldn't agree about the filtering part. While excessive sharpening can  make it look crap, actually filters like RemoveGrain and Deen or MSmooth can reduce the file size (if used properly, or the encodes will suck!). Only sharpeners and grain addition filters increase file sizes. Anyway I don't get the funda of sticking to the source. I agree using too many filters is a waste of time (literally waste of time too- encoding takes longer ), but I prefer a slight bit of sharpening and some colour tweaking.
When I said we need filters, I was talking of BTOOOM (or Ergo Proxy or some such dark ones). In episode 2 of BTOOOM, say, there is a scene for a minute and a half(about 16th to 17th min) where there are loads of blocks. I got the same when I encoded, so I had to add filters, and of course increase the psy to over 1.0 to retain the grain.
I agree it is subjective but in my opinion, if adding light amount of filters can make the encode look better, then why not?

And what about zones? I know it is slightly painful to find out the frames and all initially, but you can reduce the overall file size. It has come in useful at dark regions, where I added a high psy and low crf zone. Things like VAFE have also made zones easier.
Anyway thank you for replying (And now that I see it, Shingeki no Kyojin does seem to have different settings from what I mentioned above.) and sorry if I was wrong.

EDIT: That screenshot compare site seems pretty good. I added a couple of comparisons.
http://screenshotcomparison.com/comparison/37187
http://screenshotcomparison.com/comparison/37189
The second one was what I got at 20 crf, with and without gradfun mod(with grain strength as 2.0). Of course file size is higher, but it does not matter to me if quality is kept. In the first one, I tweaked the saturation a bit and used a warpsharpener. I prefer the encode over the source.

PS: always about The Idolmaster, I didn't notice before but some episodes are even encoded at crf 27, like ep. 20 and no filters of any kind were used, as I can see.

I like many of your encodes. Sorry if what I wanted to say came out wrongly. Anyway thank you for sharing all these encoded files despite all the recent problems with funimation. I appreciate all that you are doing.

Thanks a lot. I started encoding with Cyko in fact, not even MeGUI! I still use VAFE, for convenience, and I wouldn't claim to be an expert but it has always been my dream to encode and share anime I've encoded, after I finish studying. Don't know if there'd be scope here by then...

Just like you said, there might be more experienced people out there at other places but they don't give a damn about what others have to say. That's why I like this site a lot and hope you keep up the good work!

(reposting it here since i cant reply to zeust's thread)

RanDom wrote:I don't know how it works with negative values, I mean from what I've read so far the x264 deblock basically adds grain near the borders of the blocks to create an optical effect (I guess higher value = more grain) but at times I think that it also reduces the sharpness a lot, or at least it looks like that on my crappy 768p screen (and I don't have a lot of confidence in my eyesight too). By any chance, do you know how it works when negative values are applied or how does it affect the overall quality?

i'm not sure if Rand already had answers to this already but i've chanced upon a good explanation when i was reading over dooom9. i'll just post it here since this could help out other encoders as well, especially with the influx of the new ones. This is what mp4guy has to say:

Spoiler:

in essence, negative values help retain details by smoothing lesser artifacts/blocks while higher values tend to smooth more.

yep I read that a long time ago, really it doesn't say much about the negative values but, never mind.

Same here. Read it long ago or I wouldn't be encoding. But thanks anyway
Anyway I'm now taking a basic course on image processing so that'll probably help me learn more of the theory behind all this.

Some useful stuff in case anyone wants to learn about filters.

http://www.aquilinestudios.org/avsfilters/

Of course whether you want to use filters or not is personal choice, but it can help in compression and in cases better looking images at the cost of slowing the encode..

Man, such a loveable topic and I missed it.
Anyway, it made me realize two things.
1- Dude, for how long you're trying to get decent encode of BTOOOM? Yesterday you mentioned it too in my MPC-HC guide.
2- Like RanD said,I don't think that we use same parameters, especially recently... As I said once: CRF and AQ-strength affect the quality most and encoders here do use different values for different series.
_________________________
Lastly, a piece of info...
IMO, zones are not very user-friendly because you need to keep track of frames and then apply them, so it might be good for one or two encodes, but when you are aiming for more encodes like batch encodes, you tend to get tired of it and ditch it eventually.
However, there are two alternatives.
- Use AQ-Strength and CRF carefully in the first place.
- Use AQ-Sensitivity-curve. (That's another parameter that changes AQ strength according to some certain value given by user.)
_________________________
Hope it helps.

LOL I was done with that thing but then I saw your guide that day so I thought I'll bring it up again as I had that doubt and I thought you might have some suggestions to improve

Sorry that criticism was not meant for you.. Hope you didn't get offended. Just was trying to suggest some things.

sjoe92 wrote:LOL I was done with that thing but then I saw your guide that day so I thought I'll bring it up again as I had that doubt and I thought you might have some suggestions to improve

Sorry that criticism was not meant for you.. Hope you didn't get offended. Just was trying to suggest some things.

Not at all. I didn't even read your entire post. XD