Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion. Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules. |
14th April 2017, 02:55 | #41 | Link | |
Join Date: Mar 2006
Location: Barcelona
Posts: 5,034
|
Quote:
I prefer to do my research first. Thanks to the internet, this is a lot easier than 30 years ago when I had to go to the library on a regular basis. Lastly, there is a difference between occasionally getting something wrong and writing stuff that consistently contains 70% nonsense and 30% padding.
__________________
Groucho's Avisynth Stuff Last edited by Groucho2004; 14th April 2017 at 23:23. |
|
14th April 2017, 04:58 | #42 | Link |
Registered User
Join Date: Mar 2011
Posts: 4,823
|
Is there a type of Xvid I've so far not discovered, because I'm a little confused about the claims that Xvid is as good or better than x264. Well.... I guess it can be, but the default h263 matrix throws away plenty of fine detail, (for "old hardware" compatibility you possibly need to use it), and if memory serves me correctly Xvid's B-Frames are somewhat average quality. It's also more prone to "blockiness" in flat areas, no matter how much bitrate you give it..... at least in my experience.
And Xvid has no quality based single pass encoding method as x264 does, so it's either pick a bitrate and hope for the best, which is far from ideal (remember when every AVI was 350MB or 700MB etc?) or run a compression test before encoding. Mind you the mpeg matrix is better than h263, and some of the custom matrices (from memory) do a good job of retaining detail, but that's generally reflected in the bitrate. Not that I think anyone has mentioned the matrix they use with Xvid, although a custom matrix could hurt hardware compatibility. Is there a name for the "hazy" compression artefacts around subtitles or sharpish edges that I recall were inevitable for Xvid but far less of a problem for x264? I ran a few little sample encodes just to see. It's a 1080p source downscaled to 960x396 (Spline36) for encoding, then upscaled to 1080p again on playback by MPC-HC. To find the bitrate for Xvid's maximum quality I ran a CQ2.0 encode (Home Theatre profile, one B-frame, VAQ enabled) then I used the resulting bitrate for 2 pass encodes (1050kbps). For x264 I used the default settings (no tuning etc). I assume those claiming Xvid is wonderful must at least be using the mpeg matrix. These be thumbnails. Click for the full size version. Source x264 Xvid h263 matrix Xvid mpeg matrix Last edited by hello_hello; 14th April 2017 at 06:11. |
14th April 2017, 08:27 | #43 | Link | |
Registered User
Join Date: May 2006
Posts: 3,997
|
Quote:
It is also an explanation for some "undersizing" issues: The XviD encoder is saturated and doesn't waste extra bits for no benefit. |
|
14th April 2017, 09:03 | #44 | Link |
Registered User
Join Date: Mar 2011
Posts: 4,823
|
Yeah the way I understand it, a long time ago the quantizers were changed so the minimum quantizers could be set to "1" to allow the encoder to spend extra bits when required, to help prevent undersized files, but as far as I know it does nothing to increase the quality. I don't know what it does with them, but I'm pretty sure CQ2.0 is maximum quality.
I remember back in the AutoGK days I'd occasionally specify too high a bitrate for one reason or another and end up with an undersized file, and AutoGK was very good at outputting the requested file size as a rule. If you run a single pass, "100% quality" encode using AutoGK, you get a CQ2.0 encode, unless my memory is letting me down. 75% quality is CQ2.67, from memory. Last edited by hello_hello; 14th April 2017 at 09:12. |
14th April 2017, 09:53 | #45 | Link |
Registered User
Join Date: Mar 2011
Posts: 4,823
|
Now I'm wondering if I was wrong when I said a minimum quantizer of "1" doesn't improve quality. I'm not sure if I'm interpreting the Xvid stats correctly, but for the first encode I specified a bitrate way higher than the encoder would be able to achieve.
For the second encode I specified a more sensible bitrate. If nothing else, maybe forcing a minimum quantizer of "1" by specifying a really high bitrate will effectively improve the B-Frame quality, if the stats are anything to go by. I probably should re-run my earlier test encodes at a much higher bitrate to see if the quality does improve. Last edited by hello_hello; 14th April 2017 at 09:56. |
14th April 2017, 10:28 | #46 | Link |
Registered User
Join Date: May 2006
Posts: 3,997
|
Yes, the quantizer range is 1 to 31, with 1 being generally considered very inefficient, means no to little quality improvement at "exploding" file size. Hence the more practical value of 2 for "maximum quality".
Two more parameters to play with for quality optimization for single-pass encodes are BRatio and BOffset which determine the value of the actual B-frame Quantizer from the quantizer values of the adjacent I or P frames. Oh yeah, long time ago ...... Edit: I just found in my old notes that my recipe was for BRatio = 1.65 and BOffset = -0.25 (XViD defaults 1.5 and 1 respectively). But don't quote me on that...... Last edited by Sharc; 14th April 2017 at 10:56. |
14th April 2017, 10:32 | #47 | Link |
Registered User
Join Date: Mar 2011
Posts: 4,823
|
Looks like I'll have to do an about-face. I tried again using the h263 matrix while specifying a bitrate of 2050kbps. For some reason I actually got 2900kbps, maybe because it was a short sample, but the detail retention definitely improved. In fact I think this time the h263 matrix retained about the same amount of detail as the x264 encode, although at nearly 3x the bitrate you'd hope so.
Edit: For future reference I ran the encodes again with Xvid's overflow controls set to 10% and that way it hit the target bitrate of 2050kbps. It still wasn't doing quite as well as x264 in respect to fine detail retention even though it was now only twice the x264 bitrate, but it wasn't much worse over-all as the 2900kbps encodes when the overflow settings were 5% and it overshot the target bitrate of 2050kbps by quite a lot. Xvid h263 matrix, 2900kbps So I might have to eat my words about Xvid not being able to retain the same amount of detail as x264, assuming you let it use an extremely high bitrate (compared to x264), but even so, I'm not sure I'd be using a less efficient encoder unless I had a particular reason, such as hardware player compatibility. Last edited by hello_hello; 19th April 2017 at 05:03. |
14th April 2017, 22:45 | #48 | Link | ||
Registered User
Join Date: Feb 2002
Location: California
Posts: 2,691
|
Quote:
Quote:
I have no way of knowing your history with every member on this forum. All I know is what I read in that post. Wow, it is obvious that it doesn't take much to cross you. What the heck did I do over there? I think my history in both forums is one of helping people rather than engaging in this sort of nonsense. |
||
14th April 2017, 23:03 | #49 | Link | ||
Unavailable
Join Date: Mar 2009
Location: offline
Posts: 1,480
|
<OFF-TOPIC>
Quote:
"Almost immediately" is what you say. Why you wanted to ignore the paragraph below: Quote:
But again: ¿why the hell should I tolerate when someone miswrites my name? BTW, I also don't like when people call me "Marisa" instead of «Marsia». Over now. </OFF-TOPIC> |
||
15th April 2017, 00:45 | #50 | Link | |||
Formerly davidh*****
Join Date: Jan 2004
Posts: 2,493
|
Quote:
Quote:
That you blocked Mick because he misspelled your name was the only plausible conclusion I could come to, as well. |
|||
15th April 2017, 01:46 | #51 | Link | |
Registered User
Join Date: Mar 2011
Posts: 4,823
|
Quote:
Defaults: BRatio 1.65, BOffset -0.25 Last edited by hello_hello; 17th April 2017 at 16:02. Reason: spelling |
|
15th April 2017, 01:52 | #52 | Link | |
Unavailable
Join Date: Mar 2009
Location: offline
Posts: 1,480
|
Quote:
Another troll for my ignore list. Last edited by Midzuki; 15th April 2017 at 01:55. Reason: edit |
|
15th April 2017, 12:08 | #54 | Link | |
Unavailable
Join Date: Mar 2009
Location: offline
Posts: 1,480
|
Quote:
Yes, some low-IQ people decide to harass me because of something THEY THINK that happened, and then I have no right to complain, otherwise I am the one to be called "troll". Your logic is amazing. The logic of the emotional manipulators. The truth is, the only trolls left are the ones who claim to be "rational". |
|
15th April 2017, 12:40 | #55 | Link | |
Registered User
Join Date: May 2006
Posts: 3,997
|
Quote:
|
|
15th April 2017, 17:37 | #56 | Link | |
Registered User
Join Date: Mar 2011
Posts: 4,823
|
Quote:
Edit: Whoops, my bad. Thinking about it, I may have used BOffset = -2.5 instead of -0.25, being an idiot. I'll try again tomorrow. Last edited by hello_hello; 15th April 2017 at 17:43. |
|
16th April 2017, 11:45 | #57 | Link | |
Registered User
Join Date: Mar 2011
Posts: 4,823
|
Quote:
I wouldn't even attempt to pick a quality winner at this stage as the source quality isn't all that high to begin with. I'll probably try again later with a much higher quality source. I can see frames are encoded differently, but at the moment they're mostly only a little different rather than better or worse as such. Defaults: BRatio 1.65, BOffset -0.25 |
|
Thread Tools | Search this Thread |
Display Modes | |
|
|