Please confirm my understanding of x264 encoding settings

[gilgamesh-au]

I’ve done XviD conversions previously but I’m just starting to use x264 (which is why I’m posting here) and I would appreciate it if someone could verify my understanding of some of the settings.

My main objective is to have high quality while also keeping the file size down. It doesn’t matter to me how long the encoding process takes and I’m quite happy to leave my PC on for however long it takes. I will primarily be using PAL video as my source of the material (720x576 25 fps).

I will be using two pass encoding. Is there any effect on the quality or size of resulting video by using quick/turbo on the first pass vs normal speed on the first pass?

I plan on using the High profile. For my PAL source I would use Level 3.x. I believe that encoding at a higher level will “not” improve the quality or reduce the file size. But encoding at the unnecessary higher level will in fact mean that the player used for the resulting video will need to be capable of the higher level, potentially reducing the number of players that could handle the video. Are both of these correct?
I would be using the “Slower” preset. As I mentioned I don’t care how long the encoding takes but I understand that using the Slower preset will improve the quality or reduce the size of the resulting video (or both).

I use Gordian Knot (because I’m familiar with it) and x264vfw. I realise that the vfw versions are not official but the one I have is a recent build of 264 (b2538).

Thanks

[jagabo]

I will be using two pass encoding.

2-pass encoding doesn't deliver better quality, it delivers a known file size. When you use 2-pass encoding you know the size of the resulting video but not the quality. When you use CRF encoding you know the quality of the resulting video but not the file size. When the two methods deliver the same file size the quality is nearly indistinguishable.

Is there any effect on the quality or size of resulting video by using quick/turbo on the first pass vs normal speed on the first pass?

A slow first pass delivers a little higher quality.

I plan on using the High profile. For my PAL source I would use Level 3.x. I believe that encoding at a higher level will “not” improve the quality or reduce the file size.

Higher levels deliver better quality because more h.264 features are enabled.

But encoding at the unnecessary higher level will in fact mean that the player used for the resulting video will need to be capable of the higher level, potentially reducing the number of players that could handle the video.

Yes.

I would be using the “Slower” preset. As I mentioned I don’t care how long the encoding takes but I understand that using the Slower preset will improve the quality or reduce the size of the resulting video (or both).

Yes, but the quality difference between slow, slower, veryslow, and placebo is very small. The difference in encoding time is not.

[hello_hello]

I plan on using the High profile. For my PAL source I would use Level 3.x. I believe that encoding at a higher level will “not” improve the quality or reduce the file size. But encoding at the unnecessary higher level will in fact mean that the player used for the resulting video will need to be capable of the higher level, potentially reducing the number of players that could handle the video. Are both of these correct?
I would be using the “Slower” preset. As I mentioned I don’t care how long the encoding takes but I understand that using the Slower preset will improve the quality or reduce the size of the resulting video (or both).

Left to it's own devices it seems x264 sets Constrained Baseline Profile, Level 3.1 for DVD type resolutions, if you don't specify something else (<-- Incorrect. Corrected in a later post). I always specify High Profile, Level 4.1 myself as it's pretty widely supported these days (maybe with the exception of some cheaper portable devices).
Of course if you use level 4.1 it doesn't mean the video can't still conform to a lower level, in which case you could change the level at a later date to keep a player happy if need be. If the level 4.1 video doesn't conform to 3.1 though, you probably still wouldn't be able to play it using a device that required 3.1 even if you changed the level, but then it probably means the video was encoded differently due to using a higher level originally, which might effect the quality/file size. I can't say I've experimented with different levels much myself.
H264 Level Editor

I use Gordian Knot (because I’m familiar with it) and x264vfw. I realise that the vfw versions are not official but the one I have is a recent build of 264 (b2538).

When you've got some time you might want to give MeGUI a test drive (it has a bit of a learning curve, although if you use Gordian Knot it shouldn't be anything too foreign).

And consider CRF encoding. Encoding for quality, rather than file size, is all the rage these days.

[sneaker]

Left to it's own devices it seems x264 sets Constrained Baseline Profile, Level 3.1 for DVD type resolutions, if you don't specify something else.

No.

[hello_hello]

Left to it's own devices it seems x264 sets Constrained Baseline Profile, Level 3.1 for DVD type resolutions, if you don't specify something else.

No.

It must change according to the x264 speed preset being used. I checked some test encodes I made yesterday using the fastest x264 speed preset and that's what was set. Constrained Baseline Profile, Level 3.1.

I just tried again using the default x264 settings at 720x576 and it used High Profile, Level 3.
For a 960x544 resolution it was High Profile, Level 3.1

I guess it changes the profile and level according to the number of B and reference frames specified, or according to the speed preset being used, as well as the resolution, or something along those lines.

[sophisticles]

And consider CRF encoding. Encoding for quality, rather than file size, is all the rage these days.

Not by people that encode video professionally; it's "all the rage" for people that take a commercial Blu-Ray and decide that "making a backup" means butchering the quality by cropping, resizing and re-encoding the source to a lower bit rate.

I don't wish to go 12 rounds with anyone but let's be honest with the poster and ourselves.

[manono]

it's "all the rage" for people that take a commercial Blu-Ray and decide that "making a backup" means butchering the quality by cropping, resizing and re-encoding the source to a lower bit rate.

Since the OP's source is 720x576 I doubt he's doing as you claim.

Not by people that encode video professionally

I suppose I'm what might loosely be termed as a 'professional' since many of the videos I encode using CRF settings are for being monetized on YouTube. I encode them that way after taking lousy tape and DVD sources and doing what I can to improve them before then making MP4s for YouTube and DVDs for myself and others.

...but let's be honest with the poster and ourselves.

Yes, let's. You're full of it. There are a number of reasons one might encode using CRF that are outside the narrow range as defined by you.

[_Al_]

Left to it's own devices it seems x264 sets Constrained Baseline Profile, Level 3.1 for DVD type resolutions, if you don't specify something else.

I have similar experience that x264 set levels by the book, it takes in regard buffer settings also, perhaps other things as shown on that table. If reasonable settings including buffers and preset medium is set, encoder sets level 3.1 for SD video, if you set buffers to 40.000 it sets level 4.1, for SD video, you set double frame rate and reasonable buffers, it sets level 3.2, so it takes frame rate in an account as well following that table , etc.
IT just sets High all the time if not specified otherwise I might add.

[sophisticles]

I suppose I'm what might loosely be termed as a 'professional' since many of the videos I encode using CRF settings are for being monetized on YouTube. I encode them that way after taking lousy tape and DVD sources and doing what I can to improve them before then making MP4s for YouTube and DVDs for myself and others.

That's not being a professional, that's being a movie pirate. You take someone's content that you do not own the rights to, re-encode it, upload it to YouTube, then you try and turn page hits into a revenue stream.

I would have more respect for you if you just posted your pirated garbage to a torrent site or set up an FTP and charged a connection fee.

But more importantly it proves that even by movie pirate standards you are substandard because you take a previously compressed file, re-encode it, then upload it to a site that is known for re-compressing with atrocious settings.

Congrats, you are incompetent on a level that humbles mere mortals.

You're full of it.

At least I'm not some half baked movie pirate that tries to scrape together a few measly dollars by stealing other people's work.

Get a job!

[_Al_]

So professionals encode 2pass but amateurs, pirates and low lives encode CRF? It is so stupid that it is funny.

[ndjamena]

Presumably he'd be doing something a little more complicated than simply posting whole movies/episodes pell-mell on youtube.

[hello_hello]

And consider CRF encoding. Encoding for quality, rather than file size, is all the rage these days.

Not by people that encode video professionally; it's "all the rage" for people that take a commercial Blu-Ray and decide that "making a backup" means butchering the quality by cropping, resizing and re-encoding the source to a lower bit rate.

I don't wish to go 12 rounds with anyone but let's be honest with the poster and ourselves.

While we're being honest with ourselves, did you at least admit to yourself you couldn't get another encoder to match x264's CRF encoding at the same bitrate, or are you still working on those samples?

Did you go to the same school of generalising as newpball?
Did you notice the OP asked about re-encoding DVDs and not authoring a "professional" Bluray disc where you might as well throw as much bitrate as you can at the problem because there's room on the disc?
I've re-encoded lots of Bluray video at a lower bitrate, so I know first hand your claim that doing so means "butchering" the quality isn't always correct. Often, it's even possible to improve the quality, even when resizing it, because those who encode video "professionally" often appear not to be very good at it.
I'll happily crop the black bars when I encode, and if ever I own a wide screen TV I'll happily watch my widescreen encodes without unnecessary black bars. You'll probably be able to reminisce over the good old days of watching widescreen video on a 4:3 DVD displayed on a 16:9 TV. Seriously? Cropping is butchering? Okay...... and you went to the same school of don't resize your video as newpball too? Did you also major in ignoring evidence when it doesn't suit you while minoring in the art of making excuses when it's too hard to ignore?

I don't want to go 12 rounds with anyone either but lets have a reality check, offer sensible advice and refrain from arguing about it unnecessarily.

But more importantly it proves that even by movie pirate standards you are substandard because you take a previously compressed file, re-encode it, then upload it to a site that is known for re-compressing with atrocious settings.

Congrats, you are incompetent on a level that humbles mere mortals.

You're full of it.

At least I'm not some half baked movie pirate that tries to scrape together a few measly dollars by stealing other people's work.

That's a fairly unashamed display of ignorance. For someone who didn't wish to go 12 rounds you came out swinging blindly. Next you'll be ranting and raving about his income because he spent money building a computer.....

[_Al_]

Presumably he'd be doing something a little more complicated than simply posting whole movies/episodes pell-mell on youtube.

That is not the point, even if he did, by sophisticles measures, he is pirate and therefore if he makes money, it does not follow into his "PRO" category.
Pro's suppose to encode into given "bucket", ..., if you do not encode into size - given bucket, playing lottery, having video shorter than one hour and thirty minutes or so - you are not a PRO.

[hello_hello]

IT just sets High all the time if not specified otherwise I might add.

In case you missed it I added some more information in a new post, correcting my previous one, but it did use Constrained Baseline Profile and Level 3.1 for the 960x544 test encodes I ran the other day using the ultra fast x264 speed preset.

[ndjamena]

I have a bunch of DVDs encoded by pros... as far as I can tell someone sends them the source material, they shove the video into a machine and the machine applies whatever filters happen to be pre-programed into it irrespective of what's actually in the video and whether the filters are necessary or even completely destructive.

[hello_hello]

I have a bunch of DVDs encoded by pros... as far as I can tell someone sends them the source material, they shove the video into a machine and the machine applies whatever filters happen to be pre-programed into it irrespective of what's actually in the video and whether the filters are necessary or even completely destructive.

And then there's the aspect ratio........
I encoded some old 4:3 DVDs a while back and for a couple of the episodes the opening credits were 16:9 and stretched instead of 4:3 and I'm sure I've encoded plenty that were ITU but in places they needed to be stretched even wider for the correct aspect ratio. You'd think if you're encoding episodic DVDs and the video under the opening credits is the same each time it wouldn't be too hard to keep them looking the same and not blow out the whites and crush the blacks. Speaking of which, if video needs to be converted from full range to limited levels how is it that quite often bits of it (sometimes just a scene or two, sometimes whole episodes) appear to have been converted twice? If you happen to have watched the PAL DVDs of the re-imagined Battlestar Galactica you may have noticed "space" is dark grey in the opening credits of a couple of episodes in season three (I think) instead of black like it should be. Professionals.....

[ndjamena]

As far as I can tell the Australian releases of the Transformers Unicron Trilogy were put through two separate blended frame rate conversions to get them to PAL. I managed to return the ones I hadn't opened yet back to the store and them ordered the American releases from Amazon.com. Thankfully the replacements were mostly unblended progressive, they had been through a rather terrible ITVC process but they're the best I'm going to get and far, far better than the home grown editions.

[sophisticles]

So professionals encode 2pass but amateurs, pirates and low lives encode CRF? It is so stupid that it is funny.

What's stupid, and laughable, is allowing a program to determine quality of the produced file when said program has been marketed for years around the sales pitch that a program can't be trusted to determine quality.

The same people that promote and advocate the use of CRF are the same people that will tell you that PSNR and SSIM can't be trusted to accurately represent quality, yet they trust the encoder to determine quality based on a different contrived metric.

Now that is truly stupid.

[ndjamena]

Using CRF doesn't tell x264 to determine quality, it tells x264 how far to go using metrics it was going to use anyway. It's impossible to know how much bitrate will be necessary to encode any particular video to any particular "quality", it's also impossible to know what level of PSNR or SSIM or CRF will be noticeable for any particular video either. There's no absolutes here, you just do your best and x264 CRF is more accurate than guessing what file size is necessary for a video by simply watching it.

[_Al_]

CRF

There was a huge thread about that, with you as a participant, everything was said and explained already, advantages, disadvantages but you do not accept it for some reason, no point to go at it again. Encode 2pass, even 20 min video that goes on BD or DVD (not this topic though) or just for whatever, your choice, if it makes you feel it is going to look better.

[sophisticles]

Using CRF doesn't tell x264 to determine quality

That is exactly what it does, you are choosing an average quantizer that represents an average quality.

[ndjamena]

Setting a bitrate sets a quality as well, as does setting a file size... what's your point?

[manono]

That's not being a professional, that's being a movie pirate. You take someone's content that you do not own the rights to, re-encode it, upload it to YouTube, then you try and turn page hits into a revenue stream.

You're still full of it. I monetize public domain films and film songs. As you well know, if I didn't have the right to monetize the material in question the real copyright owners would step in and keep me from monetizing it. As it is I still have problems with liars and thieves stealing the public domain videos on which I work so hard. You should be very careful before accusing someone of something for which you have no proof at all. I've never used a torrent site in my life to either upload or download videos.

Nice try at diverting attention away from the question of why one might use CRF encoding.

[hello_hello]

So professionals encode 2pass but amateurs, pirates and low lives encode CRF? It is so stupid that it is funny.

What's stupid, and laughable, is allowing a program to determine quality of the produced file when said program has been marketed for years around the sales pitch that a program can't be trusted to determine quality.

The same people that promote and advocate the use of CRF are the same people that will tell you that PSNR and SSIM can't be trusted to accurately represent quality, yet they trust the encoder to determine quality based on a different contrived metric.

Now that is truly stupid.

Nobody claims CRF is perfect. Show me a post where someone's stated they trust it to produce exactly the same quality for a given CRF value every time. Just one. Otherwise I'll have to assume you took the "inventing people making imaginary claims" class with newpball to give yourself an argument to contradict.
Sometimes for a similar perceived quality you need to use different CRF values, but much of that is thanks to 8 bit encoding and banding etc. And noisy video sometimes requires a lower CRF value than clean video because it's much easier to see badly encoded noise than minor compression artefacts if the video is clean. At least that's how I tend to see it, but everyone has a their own perspective.
What's the alternative? Constant quantizer for a similar perceived quality at a higher bitrate? Average bitrate encoding? Disabling all the psy stuff and increasing the bitrate substantially to compensate? Where's an example of another encoder doing it better and how does it let you choose a quality? Please enlighten us.

Did either you or newpball graduate top of the class on "advanced evidence ignoring and question avoidance"? You claimed re-encoding at a lower bitrate "butchers" the video yet you happily ignored the examples I offered instead of explaining how it fails to prove you wrong. You didn't answer a single question I asked. I assume you don't expect to be taken seriously.

[Detmek]

I’ve done XviD conversions previously but I’m just starting to use x264 (which is why I’m posting here) and I would appreciate it if someone could verify my understanding of some of the settings.

My main objective is to have high quality while also keeping the file size down. It doesn’t matter to me how long the encoding process takes and I’m quite happy to leave my PC on for however long it takes. I will primarily be using PAL video as my source of the material (720x576 25 fps).

Since you care about file size, if your source have black bars, crop it. Black bars on compressed video are not pure black and will take up a solid amount of bitrate. Although, if you want to keep black bars you can crop it and add it back with Avisynth so those are solid black. Solid black needs very little bitrate to be compressed.

I will be using two pass encoding. Is there any effect on the quality or size of resulting video by using quick/turbo on the first pass vs normal speed on the first pass?

There is but you will not see it. You can encode short portion of video with slow and fast first pass, all settings same, and compare two videos. If you can see a difference then use slow first pass. If you can't, don't waste time.

I plan on using the High profile. For my PAL source I would use Level 3.x. I believe that encoding at a higher level will “not” improve the quality or reduce the file size. But encoding at the unnecessary higher level will in fact mean that the player used for the resulting video will need to be capable of the higher level, potentially reducing the number of players that could handle the video. Are both of these correct?

x264 should default to Hight profile with preset "slower". If you don't need specific level, don't set it. x264 will set it according to your settings.
Higher levels allow higher resolution, higher frame rate, maximum rate and buffere size and more refference frames. Your resolution and framerate are known. And if you don't need to limit no. of refference frames or maximum rate and buffer size don't set level. Although, setting level does not set maximum rate or buffer size, you will just limit maximum number of refference frames.

I would be using the “Slower” preset. As I mentioned I don’t care how long the encoding takes but I understand that using the Slower preset will improve the quality or reduce the size of the resulting video (or both).

If you use 2-pass encoding slower presets will not change file size, just quality and encoding time. Will the quality gain be visible for you? I don't know. But, preset "slower" should be OK for quality vs encoding time tradeoff.
If you use CRF then using slower presets will give you better quality per bitrate but it may create bigger or smaller files compared to previous or next preset.
In case of CRF encoding, first choose preset (in your case "slower") then encode a few short clips with different CRF values to see what CRF value gives you acceptable quality.

I use Gordian Knot (because I’m familiar with it) and x264vfw. I realise that the vfw versions are not official but the one I have is a recent build of 264 (b2538).

You could use x264vfw but I would suggest you to start using some more up-to-date x264 GUI, like MeGUI, Ripbotx264 or StaxRip (all based on Avisynth). Or you can go with Handbrake or VidCoder which do not use Avisynth. If you need to do a lot of cutting or use custom Avisynth filters StaxRip should do the job.

[sophisticles]

I monetize public domain films and film songs.

One of the funniest, as in sad, things I have ever read. You are the absolute worst type of "pirate", because you are basically stealing from the general public, you are taking something that is available for free and you try to make money from it.

You're kind of like the lowlifes that take free software, like a Linux distro, or an open source app and try and sell it on their website as if they own the rights to it.

Truly disgraceful and shameful, monetizing public domain films and songs, hang your head in shame.

[ndjamena]

I'm just reading the Wikipedia article on Linux, looking at all the scum of the earth who monetized the OS.

Among other it includes Intel, Nokia, TiVo, Yamaha, Samsung...

Actually, I'm pretty sure my PVR had a version of Linux under the hood.

[gilgamesh-au]

Hello,
Thanks to everyone for your responses and corrections to what I thought I understood.

Gordian knot allows me to choose a bit rate and let the software determine the file size, and doing this GK also allows 2-pass encoding. From your responses this is the wrong way to do it and is probably due to the software being so old. I will start looking into newer software as suggested.

One thing I liked about GK is that it calculated an "aspect error" when cropping the black bars, so I could then adjust the cropping to minimise the error. I haven't seen this (yet) in other software but I could still keep GK just to determine appropriate cropping to apply to the newer software.

Thanks again

[jagabo]

Gordian knot allows me to choose a bit rate and let the software determine the file size, and doing this GK also allows 2-pass encoding. From your responses this is the wrong way to do it

It depends on the situation. If you need a file of a certain size, for example you need 4.3 GiB to fit on a single layer DVD, then you want to use bitrate based encoding. In that case the encoder will give you the best quality it can (with the other parameters you specified) at that size. If you don't care about the exact size and just want to know that your video will have an approximate level of quality you use CRF encoding.

[hello_hello]

Truly disgraceful and shameful, monetizing public domain films and songs, hang your head in shame.

Obviously, you have nothing further on the topic, and once again you're in nutter mode.