Log in

View Full Version : DVD quality bitrate? (anime)


Jedd
27th August 2003, 03:47
Still trying to make a good rip. :) (just for the sake of it)
First - I had to pick up right filters, and they are mfToon() and Deen("a3d",3,10,12,20)
Second - had to set good settings for frames to handle high motion. After dozens of encodes I have selected Q:2/6/2/6, I:100/24/50
And now the final stage - pick the filesize and bitrate to make it look as good as DVD in full screen mode (or not worse then VDub preview befor compression). 20 minutes flick (640x480) 300Mb (~1700kBps) looks far from good - blocks and warp edges everywhere, (not like they are huge, no, but you can cleary see them) :mad:
Making 700Mb out of 20 minues seems to be a little pointless. So maybe I'm missing some tricks here?

Teegedeck
27th August 2003, 07:26
I don't know... Am I doing something wrong or why don't I have any problems? I don't use any filters at all or convolution3d with the animehq-preset, use exactly the same standard settings that I recommended all over the forum and get a 25 minute-episode to flawless quality at 200 megs; or to very watchable quality (slight blocks in dark hair that I can only see on my 19'' LCD, not on a CRT) at 110 MBs. So what am I doing wrong? Perhaps it's just that I don't cap quantizers, and don't filter too much. Or use trellis and qpel and B-frames. Who knows?

mf
27th August 2003, 13:44
I honestly don't know what you are doing wrong. My anime XviD settings have been like this for a long time now, and I haven't had any problems with them.

Build: Koepi's latest, dev-api-3, unstable binary.
Settings built off: Defaults.

Settings:
ME Precision: 6
Quantization: H.263
FourCC: XVID
VHQ: 1 for first pass, 4 for second (saves time)
Max I-frame itv: 300
Min I-frame itv: 1
Lumi masking OFF
Interlaced mode OFF
Quarterpel OFF
Greyscale mode OFF
Chroma motion ON
GMC OFF

Max B-frames: 2
B Ratio: 150
B Offset: 100
B Threshold: 10
DX50 B-VOP Compat ON
Packed Bitstream OFF
Debug info OFF


Min I Quant: 2
Max I Quant: 31
Min P Quant: 2
Max P Quant: 31


I-frame boost: 0
Discard first pass OFF (this is only a personal setting)
Below I-frame dist: 10
I-frame bitrate reduct: 20

High bitrate scenes: 0
Low bitrate scenes: 0
Bitrate payback delay: 250

Max bitrate: 1410064
Max overflow impr: 60
Max overflow degr: 60

Chroma optimizer ON
Trellis quant ON
Frame drop ratio: 0


That's it. I have never had these settings fail me on anime. If this doesn't work, either your filtering is wrong, or there's something with your system which is causing a bug to surface that hasn't surfaced on my systems.

Schlumpf
27th August 2003, 14:43
Originally posted by mf
...
VHQ: 1 for first pass, 4 for second (saves time)
...

Doesn't that affect the bitrate allocation? I mean... why not set it to 0 then for the first pass in order to gain even more speed?
I thought it was a golden rule that both passes should use the same settings, filters. etc.
Was I wrong?

mf
27th August 2003, 14:50
Originally posted by Schlumpf
Doesn't that affect the bitrate allocation? I mean... why not set it to 0 then for the first pass in order to gain even more speed?
I thought it was a golden rule that both passes should use the same settings, filters. etc.
Was I wrong?
VHQ4 is so slow, while it impacts the bitrate very little. XviD's ratecontrol can compensate for it.

Jedd
28th August 2003, 01:12
Well, maybe I'm just too picky here, but when I say DVD quality I mean that I should not see the difference between the same frame before and after compression. :)
Getting back to our Garry at ~1.2Mbps:
Garry, mfTooned frame before compression (http://webpages.charter.net/jedd/garry0.png)
Garry, no cap (http://webpages.charter.net/jedd/garry_nocap.png)
Garry - cap (2,6,2,6) (http://webpages.charter.net/jedd/garry_cap.png)
No caps is not an option here. I can probably drop filters, but they really make picture look better then a source (before compression at least)

Kookie
28th August 2003, 06:41
Hi,
Do you guys IVTC the anime to get the frame rate to 23.97 or just encode it at 29.97 FPS? I was scratching my head for a couple of days when my final encode came out as 29.97 FPS when I set it to 23.97 FPS in Gordian Knot. I did a search and found out that you have to IVCT it or use Force Film in DVD2AVI in order to encode it at 23.97 FPS.

Wouldn't it be better to encode the anime at 23.97 FPS reinvest the saved space into the bitrate?



This is my first post, so be gentle.:scared:

Teegedeck
28th August 2003, 07:27
Originally posted by Jedd

Getting back to our Garry at ~1.2Mbps:
Garry, mfTooned frame before compression (http://webpages.charter.net/jedd/garry0.png)
Garry, no cap (http://webpages.charter.net/jedd/garry_nocap.png)
Garry - cap (2,6,2,6) (http://webpages.charter.net/jedd/garry_cap.png)
No caps is not an option here. I can probably drop filters, but they really make picture look better then a source (before compression at least)

Sorry, Jedd, but those results you get are pitiful. This is not the normal output you can expect from XviD. I suggest you drop your filters. Maybe you like the effect of your filters, but MPEG4 certainly doesn't like blurry images (there even is a filter around that creates artificial noise because of that, you know).

Whenever someone gets extremely bad results from XviD, I'd expect to read something like 'I also tried it without filters and with XviD's default settings' at the bottom of the post. This would make it a lot easier to distinguish between real shortcomings of XviD and mere user-configuration problems.

Shalcker
28th August 2003, 08:00
As everything else, anime sources can be quite different in terms of compression. There is no universal settings. I've seen sources that got maxed out at ~150 MB for 25 minutes (without bframes, all frames at quantizer 2) and sources that had more then 1GB as first-pass size for the same 25 minutes.

You should consider revising your filter chain. Deen a3d improves compressibility a bit. With your settings it should be something from 5 to 20%. Mftoon, as averything else sharpening image, drops compressibility. When edges are sharpened you should expect more mosquito noise around them even at lowest quantizers and more (sometimes a lot more, depends on filter strength) bitrate spent on those sharpened edges too. It is often better to encode a bit blurry and sharpen at playback time if sharp image is desired.
And always IVTC anime to 23.97 fps when possible. :)

Small picture "jumps" (when frame moves a few pixels in random direction, almost unnoticeable at playback) tend to eat a lot of bitrate too.

Teegedeck
28th August 2003, 08:25
Just a note because there seems to be a die-hard misconception around:

Good compressiblity does NOT equal good quality! It's a legend from DivX3.11-days that anime must be heavily filtered and/or quants capped for anime to look good. The opposite is true.

A softened and denoised anime, compressed at quant=3 in most cases looks much worse than an unfiltered image compressed at quant=6!!! Please always try a two-pass(!) without filters first, then gradually build up a filter-chain, only adding a filter if the result after 2nd pass(!) shows an improvement. Don't cap quants!!!

mf
28th August 2003, 11:01
Teegedeck, analogue anime is NOT the same thing as digital anime. With digital anime, you only need to filter out compression artifacts, with analogue you'll have a lot more to worry about. Please believe this from an experienced encoder.

Shalcker
28th August 2003, 12:47
Originally posted by Teegedeck
A softened and denoised anime, compressed at quant=3 in most cases looks much worse than an unfiltered image compressed at quant=6!!!
I'd like to object "in most cases" but that can greatly differ depending on cases involved.

If you see that image quality IMPROVED after filtering AND it also got LOWER quantizers while encoding to target bitrate then you should certainly use filtered source.
If you see that image quality IMPROVED after filtering BUT it got HIGHER quantizers while encoding to target bitrate then you should check result and consider dropping some "sharping" filters IF you're not satisfied with final image quality (sometimes filtered image may look better to you even if it got higher quantizer).

If you see that image quality DEGRADED after filtering BUT it also got LOWER quantizers while encoding to target bitrate then you should compare it with unfiltered result encoded to SAME BITRATE and check which distortion is better for your eyes - a bit more MPEG4 artifacts -or- filter effects and a bit less MPEG4 artifacts (since sometimes unfiltered image may also look better even having higher quantizer).
If you see that image quality DEGRADED after filtering AND it got HIGHER quantizers you should drop your filters.

The problems are that image quality is highly subjective, can differ in certain parts depending on filters used, and not too many people bother to check unfiltered or partially filtered source since it's quite time consuming.
If image quality is improved after filtering AND result of first pass is lower filesize then filtered source is better in any case.
But only as long as filtered image is considered "improved". It may look awful to another person.

Angrychair
28th August 2003, 14:55
Originally posted by Teegedeck
Just a note because there seems to be a die-hard misconception around:

Good compressiblity does NOT equal good quality! It's a legend from DivX3.11-days that anime must be heavily filtered and/or quants capped for anime to look good. The opposite is true.

A softened and denoised anime, compressed at quant=3 in most cases looks much worse than an unfiltered image compressed at quant=6!!! Please always try a two-pass(!) without filters first, then gradually build up a filter-chain, only adding a filter if the result after 2nd pass(!) shows an improvement. Don't cap quants!!!

While I've seen plenty of examples of what I consider to be very poorly filtered anime (I call it sharpened soup, as they blur the shit ouf of it, then run warpsharp on it, leaving nothing but the loudest of details still present) I've found that properly filtered anime almost always looks better than unfiltered anime per meg. Now, with the right source, filtering wouldn't be necessary, but that's a rare find. (I've never found such a source, myself)

Anyway, I agree about capping quantizers, and think anime encoders should get rid of the oversharpening routine, it really doesn't look any better. You can't create real details once you've gotten rid of them.

Teegedeck
28th August 2003, 14:58
Well-said. :)

To clear it up, I'm really talking about anime, not that CG-animation stuff, that's for sure. And of course I wouldn't write this if I hadn't tried it myself. Anime encoding really isn't an obscure artform; there's much more talk about the 'oddities' of anime than actually is justified. I really can't stand some of the false information, based on susceptible tests that don't want to die out. Like, someone scans his encoding until he finds a blocky B-frame, compares it to a preceding non-blocky P-frame and concludes unwittingly that B-frames are a no-no for anime. If he just had tried encoding the whole thing again without B-frames, he could not only have found out that he'd have gotten a blocky P-frame in the same place where he found the blocky B-frame before, but also that his whole encoding looks worse without B-frames. And again, reducing the filesize (='improving' compressiblity) of your encoding by filtering is much worse than reducing it by higher quantization in XviD - much, much worse.

Another thing that I personally despise [really just paraphrasing in a long, winded way of talking what Angychair wrote, sorry] is that many people first remove all real detail from anime in order to reach better compressibility by smoothing/denoising it, and then apply a sharpener because the result has become too soft, thus loosing that compressibility again. What they get out of it is a source that doesn't compress much better but lost the look of the original. This artificial look seems to be quite popular or at least I get that impression. Of course I'm confident that people who hang around this forum are wise enough to do better than that, so this doesn't apply to you. Sorry that I used the occasion to get something off my chest. :D

Again, I'm not against slight filtering. Only if you get a picture that looks as if somebody smeared jelly all over the camera lense this is prone to give quantization errors. You can't judge anything by how the avs looks if loaded into VDub. Only the result of the 2nd pass counts. So I'd recommend to anyone to first do an un-/minimally filtered encoding as an anchor, and afterwards slowly build up your filter-chain. OK?

Jedd
28th August 2003, 15:06
Here the source (http://webpages.charter.net/jedd/ORIG.png) frame before filters, resize and compression.
So, if I drop all filters the best result I can get is the same quality-wise frame after compression. Which means I have failed to make a good looking rip. (and now have to rely on postprocessing).

Again, what's the difference between analogue and digital anime? (I guess I have the analogue case here)

Teegedeck
28th August 2003, 15:10
It doesn't look bad enough to me to justify any denoising. Actually, where is any noise in that picture?

And who said that you are obliged to make a DVD-quality encoding out of a non-DVD-quality source? For that you'd have to invent details that just aren't there. Your best bet for a subjective improvement is XviD's quarterpel mode.

Jedd
28th August 2003, 16:17
Originally posted by Teegedeck
It doesn't look bad enough to me to justify any denoising. Actually, where is any noise in that picture?

And who said that you are obliged to make a DVD-quality encoding out of a non-DVD quality source? For that you'd have to invent details that just aren't there. Your best bet for a subjective improvement is XviD's quarterpel mode.

Noise level is quite ok here, but overall impression doesn't satisfy me at all. I think I have exactly that "jelly" feeling.

So I have decided to play with filters to see what I can get out of it. mfToon() improved it to something I can live with:
mfToon (http://webpages.charter.net/jedd/MF.png). And then I just had to apply deen() to clean the blocks.

Maybe you are right and I can't improve quality with encoding. I just was hoping filters can help me with that. :(

mf
28th August 2003, 18:34
Originally posted by Teegedeck
It doesn't look bad enough to me to justify any denoising. Actually, where is any noise in that picture?

And who said that you are obliged to make a DVD-quality encoding out of a non-DVD-quality source? For that you'd have to invent details that just aren't there. Your best bet for a subjective improvement is XviD's quarterpel mode.
No anime experience alert!

Anime doesn't have much detail, and you certainly won't have to invent them. The thing with anime quality is cel stableness (no noise, just one flat area, no temporal fluctuations) and line sharpness and darkness. And quarterpel on anime won't help you. It might help compression on the slight shaking that analogue anime has, and that's it. I CLEARLY hear someone used to encoding real-life movies judging anime. I could be wrong, but if you're not experienced in anime encoding, please leave your judging to the ones that are.

Teegedeck
28th August 2003, 18:59
Of course I bow before your experience, MF, but I certainly do have some experience myself. What you describe is CG, but anime isn't CG (completely, yet). Cells aren't 100% flat coloured areas, but they do contain shades, texture, noise and all the stuff we love... If it was like you described, resolution wouldn't count, either. I wouldn't see a difference between a 720x and a 640x resolution anime. We could as well encode at 320x.

I don't know whether you have tried quarterpel yourself. No quarterpel experience alert? Don't jump to conclusions...

Be careful.

mf
28th August 2003, 19:26
Originally posted by Teegedeck
Of course I bow before your experience, MF, but I certainly do have some experience myself. What you describe is CG, but anime isn't CG (completely, yet). Cells aren't 100% flat coloured areas, but they do contain shades, texture, noise and all the stuff we love... If it was like you described, resolution wouldn't count, either. I wouldn't see a difference between a 720x and a 640x resolution anime. We could as well encode at 320x.

I don't know whether you have tried quarterpel yourself. No quarterpel experience alert? Don't jump to conclusions...

Be careful.
I don't claim them to be 100% flat, but usually a gradient or static texture is as far as it gets. Noise is one thing they don't contain, or at least shouldn't contain, and resolution does count. Backgrounds are the only things that occur often that are truly textured, but as they're (mostly and usually) static, a proper 3D smoother will leave them alone. Low resolution makes edges crude and cripples sharpness, but it's true that I can make a 320x240 source very close to a 640x480 source, with proper resizing and filtering. And yes, I've tried qpel as well, but I honestly saw no improvement (compression might have been better). It could be that there was some more noise-like stuff, but as I never encoded the same episode without qpel I wouldn't truly know. The episode in particular didn't look any better than the others without qpel. Anyway, I don't like extra noise on an anime encode anyway.

Teegedeck
28th August 2003, 19:46
Granted, old chap. I dig your comments, as long as they're not too spicy. ;) But I realize we're actually discussing things that depend a bit too much on taste in order to make general recommendations. Some will like it this way and filter as much as they deem neccessary, others will like it the other way and filter as little as possible.

mf
28th August 2003, 19:50
Originally posted by Teegedeck
Granted, old chap. I dig your comments, as long as they're not too spicy. ;) But I realize we're actually discussing things that depend a bit too much on taste in order to make general recommendations. Some will like it this way and filter as much as they deem neccessary, others will like it the other way and filter as little as possible.
Yeah, I got a little annoyed. Even then, I tried to keep my comments not too hostile :D.

Jedd
28th August 2003, 23:53
Peace. http://fool.exler.ru/sm/alc.gif

Originally posted by Teegedeck
Whenever someone gets extremely bad results from XviD, I'd expect to read something like 'I also tried it without filters and with XviD's default settings' at the bottom of the post. This would make it a lot easier to distinguish between real shortcomings of XviD and mere user-configuration problems.
Sorry, didn't post it here, here we go:
Garry - no nothing (http://webpages.charter.net/jedd/garry.avi)
Done with all default settings, no filters whatsoever, ~1000kbps. You can pause it it any place, to make sure it's not a bad B-frame I'm trying to improve.

PowerMacG4
29th August 2003, 05:55
Most of that clip is q31, and it appears to be all I/P-frames with no motion vectors. I do not see a single B-frame.

Teegedeck
29th August 2003, 07:49
Very strange. Jedd, did you by any chance select 'none' under 'motion-search precision'?

Joe Fenton
30th August 2003, 04:13
Originally posted by Kookie
Hi,
Do you guys IVTC the anime to get the frame rate to 23.97 or just encode it at 29.97 FPS? I was scratching my head for a couple of days when my final encode came out as 29.97 FPS when I set it to 23.97 FPS in Gordian Knot. I did a search and found out that you have to IVCT it or use Force Film in DVD2AVI in order to encode it at 23.97 FPS.

Wouldn't it be better to encode the anime at 23.97 FPS reinvest the saved space into the bitrate?

This is my first post, so be gentle.:scared:

This is personal experience on making backups of my own DVDs (not endorsing illegal copying here). There are two processes in making animation: hand-drawn cells and computer generated sequences. Most old anime tends to be almost completely cells with almost no computer sequences. In fact, sometimes the ONLY computer sequences are the credits at the end. Cells are recorded on film at 24 FPS; when transferred to computer, they are resynced to 23.976 FPS, then telecined to 29.97 FPS. This is then composited with the computer generated 29.97 FPS video. This makes the whole question rather fun - some of the anime is at 24 FPS, and some at 30 FPS. The whole question comes down to how much?

When saving a project in DVD2AVI, it tells you how many frames are at 23.976 (FILM) rate, and how many are at 29.97 (NTSC) rate. For a modern anime like Cowboy Bebop, the main animation is about 80% FILM until you get to the credits, which are 100% NTSC. I use this rule of thumb: if DVD2AVI reports 70% or better FILM (30% or less NTSC) OVERALL, set the option in DVD2AVI to save out the project at forced film rate (23.976), otherwise save the project without any forced film rate (29.97).

Why do this? There are two reasons: first, as mentioned by Kookie, you save bitrate. This allows more bits to be used elsewhere for better quality or smaller size. Second, motion is MUCH smoother if you use the rate that MOST of the action occurs at. If it is 70% film speed, it looks much smoother at 23.976 FPS. Some recent movie DVDs I have bought are 99.9% film rate! These SUCK at 29.97 FPS. Always use Forced Film in DVD2AVI for these.

DO NOT USE Inverse Telecine filters!!! They suck compared to using Force Film in DVD2AVI, and then NOT using a filter in VirtualDub. I have yet to find an inverse telecine filter that does anywhere near as good a job as just saving the project as Forced Film 23.976 in DVD2AVI. The only thing I have AVISynth/VirtualDub doing is a standard field deinterlace, crop, and resize.

Back to Cowboy Bebop - encoding at 23.976 makes the main part of the anime much cleaner and smoother. It looks awesome. The credits won't scroll smoothly as they are done at 100% NTSC rate, but it's a tradeoff I'm willing to live with.

manono
30th August 2003, 06:46
Bad advice, Joe. All of it, bad advice.

http://www.doom9.org/ivtc-tut.htm

Joe Fenton
30th August 2003, 07:06
Originally posted by manono
Bad advice, Joe. All of it, bad advice.

http://www.doom9.org/ivtc-tut.htm

Sorry, but that page is BOGUS NONSENSE! Emphasis on BOGUS.

I've done what they recommend on that page and it turns out like CRAP! Do it my way if you want it to look good. DVD2AVI works just fine down to 70% in every rip I've checked and certainly looks FAR superior to the recommended telecine/decimate suggestion. (Blowing raspberry at stupid author of suggested page)

I find it hard to believe the author of that page actually did what he himself recommended AND THOUGHT IT LOOKED BETTER! (Shaking head sadly)
:rolleyes:

KpeX
30th August 2003, 07:15
Originally posted by Joe Fenton
Sorry, but that page is BOGUS NONSENSE! Emphasis on BOGUS.

I've done what they recommend on that page and it turns out like CRAP! Do it my way if you want it to look good. DVD2AVI works just fine down to 70% in every rip I've checked and certainly looks FAR superior to the recommended telecine/decimate suggestion. (Blowing raspberry at stupid author of suggested page)

I find it hard to believe the author of that page actually did what he himself recommended AND THOUGHT IT LOOKED BETTER! (Shaking head sadly)
:rolleyes:

Wow. You better duck.

That page was written by three of the most respected members of this forum, and they have forgotten more about video processing than many of us (Joe) will ever know. The methods on that page are tested and proven, and the best advice IMO anywhere on the web. So what DVD do you claim came out like crap? What did you use to IVTC? Even with high percentages of film in D2A, IVTC plugins will still do a fine job.

mf
30th August 2003, 11:07
Heh, I must agree, for some weird reason, DVD2AVI's force film creates less (ALOT less - I have a particular one that Telecide totally fails on, while force film only has sporadic problems) mismatches than Telecide. It's just a pity you can't let it output at 29.97fps, so everything force film hasn't caught can be caught by some other IVTC, like Telecide.

manono
30th August 2003, 12:19
Come on, mf, you know better than that. Just to pick out his most egregious contention; if I understand him correctly, he's saying that anything under 70% Film in DVD2AVI should be kept at 29.97fps and deinterlaced. And you know as well as I do that's pure nonsense. Lots of anime, and many films as well, are encoded as interlaced (100% NTSC) after the telecine process and can be IVTC'd perfectly easily.

I couldn't care less how he encodes his own anime. But he's presenting this as sage advice to a newcomer to the site, and I can't let it pass.

mf
30th August 2003, 13:52
Originally posted by manono
Come on, mf, you know better than that. Just to pick out his most egregious contention; if I understand him correctly, he's saying that anything under 70% Film in DVD2AVI should be kept at 29.97fps and deinterlaced. And you know as well as I do that's pure nonsense. Lots of anime, and many films as well, are encoded as interlaced (100% NTSC) after the telecine process and can be IVTC'd perfectly easily.

I couldn't care less how he encodes his own anime. But he's presenting this as sage advice to a newcomer to the site, and I can't let it pass.
I'm not saying he's right with everything that he says, but for as far as I skimmed through that doom9 guide, it doesn't at all apply to anime. Ok, so I re-read it, and it might apply to it after all, but still.. ;)

Joe Fenton
30th August 2003, 20:19
Originally posted by manono
Come on, mf, you know better than that. Just to pick out his most egregious contention; if I understand him correctly, he's saying that anything under 70% Film in DVD2AVI should be kept at 29.97fps and deinterlaced. And you know as well as I do that's pure nonsense. Lots of anime, and many films as well, are encoded as interlaced (100% NTSC) after the telecine process and can be IVTC'd perfectly easily.

I couldn't care less how he encodes his own anime. But he's presenting this as sage advice to a newcomer to the site, and I can't let it pass.

It's better advice than given on that page. Even mf agrees after doing encoding tests. I struggled for weeks trying to make Cowboy Bebop and Tenchi Muyo OVAs work with the suggestions on that page. It always came out as a wavering mess. Forced Film 23.976 was the way to go on both. I have also had better luck using Forced Film on live action movies as well. I won't say it handles EVERYTHING better, but it's a better starting place than what's recommended. Sorry, but that's what my experience has taught me. I try every new method of deinterlace / inverse telecine that comes along. So far, it's either Forced Film or plain field deinterlace.

It's not sage advice, it's PRACTICAL advice from someone who has tried everything recommended and seen with his own two eyes what came out best. My advice is to save a five minute clip from DVD2AVI as Forced Film and non-Forced, then try the various IVTC methods. Pick the one that looks best. So far, it HASN'T been IVTC. Be sure to select that five minute clip from an area where you have the highest mix of FILM and NTSC.

Remember, practical advice beats sage advice anytime.

mf
31st August 2003, 00:27
Originally posted by Joe Fenton
Even mf agrees after doing encoding tests.
EVEN mf? Am I getting some kind of reputation here? :p I thought I was just an annoying guy who breaks the forum rules :D.

Aktan
31st August 2003, 03:10
@Joe Fenton

For Cowboy Bebop, try NO Force FILM and the avisynth filter IT.dll. Its perfect. Now Tenchi Muyo OVA as at 95% FILM anyway so of course Force FILM will work :p .

IT.dll seems to work very well for anime after year 2000. This is just what I have noticed

The People's Elbow
31st August 2003, 20:47
Anybody able to remember the best (and most times heavily time consuming) method?
I'm talking about manual ivtc in tmpeg!
Still the only way to 100% quality...

greetz,
Elbow!

Teegedeck
31st August 2003, 22:05
Originally posted by Teegedeck
Very strange. Jedd, did you by any chance select 'none' under 'motion-search precision'?

Nothing from Jedd since the 28th - does that mean 'yes'?

Sorry for being on-topic. :p

Jedd
2nd September 2003, 02:21
Originally posted by Teegedeck
Nothing from Jedd since the 28th - does that mean 'yes'?

Sorry for being on-topic. :p
You are wery welcome to be on-topic, Teegedeck.
Sorry for not answering your question, it was a long weekend here, so I was paddling in Vermont far from internet, computers and codecs. :rolleyes:
As for motion search - it's 6-Ultra High, of course. :sly:

Jedd
5th September 2003, 01:26
Actually, I'm thinking - maybe I should decrease motion search, maybe it treats the whole movie as fast motion. There are no 2 same frames even in the still scenes - they all shakes.

unmei
5th September 2003, 03:07
not the good countermeasure IMHO, remember this isn't some divx 3.11 with fast and low motion codecs to switch between. I dont think decreasing motion search will help you in this.
(hell its even on max as default isnt it? maybe the 1-5 are only tradeoff for real-time encoding :)
..but if its shaking all the time, using Qpel were probably inappropriate (i use that all the time, but i probably would turn it off in this case)

Joe Fenton
5th September 2003, 03:49
I'm probably mistaken, but isn't TomsMoComp supposed to help videos that "shimmer" or shake slightly? You might check that out.

Jedd
5th September 2003, 18:37
Originally posted by Joe Fenton
I'm probably mistaken, but isn't TomsMoComp supposed to help videos that "shimmer" or shake slightly? You might check that out.
So, do you suggest to replace Telecide().Decimate(5) with TomsMoComp(-1,5,1).Decimate(5)?
Although I'm not sure how deinterlace filter can help with the shaking picture.. but won't hurt to try anyway, thanks.

Jedd
7th September 2003, 03:39
Ok, did several encodes - Telecide+Qpel looks the best.
TomsMoComp doesnt help with the shaking picture at all, some places making it even worse, scrambling thin lines.
Oh, my..

t3ch
9th September 2003, 15:13
Regarding some of the argument at the beginning of the thread, some anime (older ones) need to be heavily filtered. I'm not a fan of massive smoothing then massive sharpening, don't get me wrong. However older anime is pretty grainy, which needs to be smoothened (sometimes relatively heavily) for a cleaner picture. Adding a light sharpening afterward cleans up some of the damage done by the smoothing. It should look better than the dvd source. I'm not saying you can invent details (not that all anime is extremely detailed anyways :D ), but when you remove all the crappyness in the source, it sure makes a big difference. :)

Sorry I couldn't post anything helpful earlier, I had the 5 day post waiting period :rolleyes: Looks like you got everything figured out though, so...