Good MPEG-2 Encoders out there?

[zerowalker]

Well as the Topic says, i am looking towards CCE, but not sure if it´s good?

And what is the Highest Bitrate (Constant) you can achieve while still being able to play flawlessly on a DVD Player?

(Can say that 9.5k isn´t;P)

PS:

Hope this topic is okay, know there isn´t an Encoder that is Good in all terms, so was more a matter of speak:)

[SassBot]

The best choices are CCE and HCEnc. It all comes down to if you want to spend money or not.

[zerowalker]

How is x262?

And well depends on, is CCE better than HCEnc?
If so, in what stuff, or is it overall?

[kolak]

For progressive content CCE-SP3 is still unmatched. SP3 does not like that much interlaced content. Carbon Coder has also good mpeg2 encoder, specially for interlaced sources.

[zerowalker]

Ah, my content will mostly be Interlaced (VHS content, i think it´s 50i).

[SassBot]

How is x262?

Nowhere near production ready, but highly promising.

[zerowalker]

Nowhere near production ready, but highly promising.

Okay, cause i have tried it, and well it´s not Very Good, but it´s not Bad, though i am not a encoder professional.

But it would be awesome if more work got a hold on it, atleast they have interlaced working now:)

[TheSkiller]

I use CCE SP3 a lot with interlaced videos for DVD and found it to be very good, I can't say it's particularly much worse with interlaced video than progressive. :)
However yes, with really really hard to encode stuff, and especially at lower bitrates, Carbon Coder (ProCoder) is better due to it's motion search and ability to encode per field rather than per frame, but the disadvantage is it has it's annoying built-in filters (like a denoiser) that cannot be switched off. Depending on the video, at least for me, this is pretty much unacceptable. And for progressive videos I'd never suggest ProCoder, because that's where none of it's few benefits would come in.

HCEnc is very good as well overall (and free!). I just prefer CCE SP3 because, like Kolak said, it is unmatched with keeping fine grain. Basically, if you disable all of it's low-pass filters, the encoded result is almost 100% identical to the source (for the eye) if the combination of bitrate and motion/detail allows. CCE SP3 is especially good at keeping gradients and dark areas nice. It will do great for interlaced VHS transfers (but only with high bitrates), I have actually used it for that as well, at 9200 kbit/s CBR. At medium bitrates (4000-8000), I would suggest HCEnc if it's a demanding interlaced video. Home videos are usually demanding because of the constant camera movements.


The highest bitrate for the video alone depends on the audio it will be muxed with. Maximum allowed combined bitrate of video+audio+subtitles is 9800 kbit/s.
So if you have a single 256 kbit/s audio stream for example, the theoretical maximum would be 9544 kbit/s for the video. But better don't use this value, as it's the very edge of what's allowed. I have done plenty 9200 kbit/s CBR encodes (audio 256 or less) and on many various players I have never had a problem. If I do VBR encodes I usually use a maximum of 9200 as well. Maybe 9300 or 9400 would work just as well but I'd rather leave a 200-300 kbit/s margin to the absolute max.

[zerowalker]

I use CCE SP3 a lot with interlaced videos for DVD and found it to be very good, I can't say it's particularly much worse with interlaced video than progressive. :)
However yes, with really really hard to encode stuff, and especially at lower bitrates, Carbon Coder (ProCoder) is better due to it's motion search and ability to encode per field rather than per frame, but the disadvantage is it has it's annoying built-in filters (like a denoiser) that cannot be switched off. Depending on the video, at least for me, this is pretty much unacceptable. And for progressive videos I'd never suggest ProCoder, because that's where none of it's few benefits would come in.

HCEnc is very good as well overall (and free!). I just prefer CCE SP3 because, like Kolak said, it is unmatched with keeping fine grain. Basically, if you disable all of it's low-pass filters, the encoded result is almost 100% identical to the source (for the eye) if the combination of bitrate and motion/detail allows. CCE SP3 is especially good at keeping gradients and dark areas nice. It will do great for interlaced VHS transfers (but only with high bitrates), I have actually used it for that as well, at 9200 kbit/s CBR. At medium bitrates (4000-8000), I would suggest HCEnc if it's a demanding interlaced video. Home videos are usually demanding because of the constant camera movements.


The highest bitrate for the video alone depends on the audio it will be muxed with. Maximum allowed combined bitrate of video+audio+subtitles is 9800 kbit/s.
So if you have a single 256 kbit/s audio stream for example, the theoretical maximum would be 9544 kbit/s for the video. But better don't use this value, as it's the very edge of what's allowed. I have done plenty 9200 kbit/s CBR encodes (audio 256 or less) and on many various players I have never had a problem. If I do VBR encodes I usually use a maximum of 9200 as well. Maybe 9300 or 9400 would work just as well but I'd rather leave a 200-300 kbit/s margin to the absolute max.

Ah well as you say, I can´t have a denoiser on aswell, for VHS sources that´s unacceptable, for many reasons when restoring quality.

Hmm don´t know what i have been using, i know i tried 9.5k, and it went well for a while, then i could stutter, so i guess the buffer went out.

But will try with 9.2k if i use 256 or less:)

And it sounds like i will use CCE, but i wonder, what are the settings to keep it interlaced?

I think i read that it De-interlace, and i don´t want that for 50i stuff, as DVD doesn´t support 50p.

[TheSkiller]

Ah well as you say, I can´t have a denoiser on aswellCarbon Coder (ProCoder) is not a strong denoiser but it tends to blur the picture a tad and make slight noise disappear in an ugly way, as if the noise "freezes" and then gets blurred if the bitrate is too low. Using the fastest settings helps with this and gives a more natural fine detail/grain presentation, the motion search is still very good at the fastest settings.



9500 Kbit/s for the video alone is very much, like I said, better use 9200. There will be no visible difference between 9500 and 9200.
If you have mono audio, you should encode that to a mono MP2 or AC3 audio with 128 kbit/s or a bit less. In that case you can rise the video bitrate by ~100 kbit/s to maybe 9300.


And it sounds like i will use CCE, but i wonder, what are the settings to keep it interlaced?Unless you specify different, CCE SP3 (and the old SP2) will encode interlaced. You should pay attention to the field order of the video (Top Field First or Bottom Field First) and tick the "Output top field first stream" checkbox in the main window in case of TFF video. In case of BFF video, leave it unticked.
I would also suggest to disable all of the low-pass filters in the "Picture" settings, they just blur the picture to make it more compressible. They are used by default.

I think i read that it De-interlace, and i don´t want that for 50i stuff, as DVD doesn´t support 50p. Of course all the mentioned encoders do encode interlaced video properly, they don't deinterlace. I think CCE SP3 has the ability to deinterlace if you want it to but who would use that? :)

[Guest]

The thread wandered seriously off-topic so I have broken out the disucssion on discs and disc burning to this thread:

http://forum.doom9.org/showthread.php?t=165677

Please stay on topic here: Good MPEG-2 Encoders.

[zerowalker]

what should GOP M and N be?
It´s PAL, i read something about 15,3 or something?

[TheSkiller]

The biggest GOP CCE will ever produce is 15 frames long (i.e. PAL DVD compliant, as long as there is no pulldown applied afterwards). That would be M=3 N=5 (which is the highest it will accept respectively).
For NTSC DVD without pulldown the GOP is allowed to be as long as 18 frames but CCE somehow cannot make GOPs longer than 15 frames, don't ask me why.

[zerowalker]

the lowest i can choose is 3M 6N for some reason?

and how do i disable Interlaced, meaning that it marks the MPEG-2 as Interlaced, even when it´s progressive?

[Biggiesized]

IIRC, PAL 50i and 23.98p NTSC should be set to M=3 N=4 whereas 29.97i NTSC should be set to M=3 N=5.

[zerowalker]

i can´t go lower than 6N

[TheSkiller]

and how do i disable Interlaced, meaning that it marks the MPEG-2 as Interlaced, even when it´s progressive?In "Picture" settings set "Picture type" to "Progressive". And if you want to flag it 100% progressive, then you need to set the "progressive sequence" flag afterwards using Restream (http://www.videohelp.com/tools/Restream) (no way to set this flag in in CCE).

Oh, and be sure to actually use your settings for encoding. It should look similar to this in the main window:
http://img7.imagebanana.com/img/1p6078aw/ccesp3settings.PNG



IIRC, PAL 50i and 23.98p NTSC should be set to M=3 N=4 whereas 29.97i NTSC should be set to M=3 N=5.Yes, that's what the CCE manual says and while it is true that these settings will produce legal GOP lengths, for PAL without pulldown this will produce GOPs which are shorter than they need to be.

For example, this is how a typical GOP in the middle of a calm scene looks like, encoded by CCE SP3 using the maximum M/N values it accepts (which are M=15 N=3).


I B B P B B P B B P B B P B B I
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Muxing this with Muxman to a PAL DVD and the log file says
longest GOP has 30 fields
So, PAL DVD is perfectly fine with M=15 N=3. :)

In CCE SP2 M=3 N=4 would produce a GOP that is 12 frames (24 fields) long. So I'd say, use M=3 N=4 only for pulldown (23.976 base frame rate). Everything else, PAL and NTSC 29.97, use M=3 N=5.

Blue stuff was edited later by me for convenience.

[Lyris]

HCEnc is very good as well overall (and free!). I just prefer CCE SP3 because, like Kolak said, it is unmatched with keeping fine grain. Basically, if you disable all of it's low-pass filters, the encoded result is almost 100% identical to the source (for the eye) if the combination of bitrate and motion/detail allows.

Couldn't have said it better. I love CCE SP3 for this reason. With a high bit rate and some content, CCE can give results that frankly don't look like MPEG-2 at times.

HCEnc is good too, but in my tests, it crunched film grain up more than CCE.

[kolak]

In "Picture" settings set "Picture type" to "Progressive". And if you want to flag it 100% progressive, then you need to set the "progressive sequence" flag afterwards using Restream (http://www.videohelp.com/tools/Restream) (no way to set this flag in in CCE).



I suspect that progressive sequence=1 is out of DVD spec.

[kolak]

M=3 N=4 would produce a GOP that is 12 frames (24 fields) long. So I'd say, use M=3 N=4 only for pulldown (23.976 base frame rate). Everything else, PAL and NTSC 29.97, use M=3 N=5.

Valid GOP size for DVD has to be <0.606 sec. 18 frame for NTSC is fine. In case of 23.976+pulldown it has to be 12 (or less).

[Mug Funky]

just throwing another mention of x262 out there - i compiled the most recent interlaced changes, and they make a BIG difference.

in my 1 test so far (on nightmare handheld canon 550D 50p that's been interlaced into 50i), it outperformed HC significantly.

i need to throw more stuff at it now.

[TheSkiller]

Valid GOP size for DVD has to be <0.606 sec. 18 frame for NTSC is fine. In case of 23.976+pulldown it has to be 12 (or less).Yes, Kolak. I never said differently, did I? :)

CCE SP2 settings: (see edit)
M=3 N=5 — max GOP length 15 frames
M=3 N=4 — max GOP length 12 frames

Edit: Ah, just noticed that this whole M/N thingy is not the same between CCE SP2 and CCE SP3. Everything I said before about the M/N values was referring to CCE SP2.
In SP3 the M value represents the max GOP length, much easier. And N adjusts the B frames.
Sorry for the inconvenience.

But there's no way to have CCE (any) produce a GOP which is 18 frames long (try it), which means one would use 15 frames for both pulldown-less PAL and NTSC.

[kolak]

Yes, Kolak. I never said differently, did I? :)

CCE SP2 settings: (see edit)
M=3 N=5 — max GOP length 15 frames
M=3 N=4 — max GOP length 12 frames

Edit: Ah, just noticed that this whole M/N thingy is not the same between CCE SP2 and CCE SP3. Everything I said before about the M/N values was referring to CCE SP2.
In SP3 the M value represents the max GOP length, much easier. And N adjusts the B frames.
Sorry for the inconvenience.

But there's no way to have CCE (any) produce a GOP which is 18 frames long (try it), which means one would use 15 frames for both pulldown-less PAL and NTSC.

No you didn't- key point is 0.606 sec as per DVD spec. Max number or frames comes from this value- not like someone has made it up.

[zerowalker]

So do you want as much b frames and as long GOP as possible:)?

[TheSkiller]

More than 2 consecutive B frames are not allowed for DVD. And if I remember correctly, you can't make CCE encode more than 2 consecutive B frames anyway. But even if it was possible, it would most likely result in inferior picture quality, too much B frames are no good thing. :)
In fact, CCE SP3 will often create P frames instead of B frames whenever it would provide better quality. Because of that it is not unusual for a GOP to look like IPBPBPPPBPBP for example.

What you can do though is restricting the maximum amount of consecutive B frames to 1. I did test this and would recommend not to do it because too few B frames are no good either in many cases. CCE SP3 regulates this pretty well on it's own. :)

The GOP length should be as long as allowed in any case of course (i.e. not breaking the rule Kolak mentioned).

[kolak]

SP3 and Carbon Coder have adaptive GOP feature, so they can change GOP length and structure as desired to achieve better quality. It works quite well, specially for fades, cross fades, fast action, eg trailers. The only difference is that Carbon Coder can create GOPs without B frames and this is not good, but only for very tiny amount of players. SP3 has special setting to avoid GOPs without B frames- it's was introduced to make SP3 streams as much players friendly as possible.

[Blue_MiSfit]

Carbon's MPEG-2 encoder is pretty good. There are actually 2 different MPEG-2 encoders in Carbon - the older Canopus encoder that has always been there, and a newer implementation of Mainconcept. They're quite similar in my experience.

To be honest, as long as we're talking pro encoders, the Elemental Server / Elemental Live stuff (NOT their consumer implementation) is absolutely amazing. I did a shoot-out of relatively difficult MPEG-2 encoding - i.e. 14mbps (for CableLabs VOD compatibility) 1080i high motion content. Between Carbon, Digital Rapids, and Elemental, the Elemental result looked the best to me by far, and was about 5 times faster than the 2 software encoders, even at fastest speed settings..

[zerowalker]

Okay:)

No, i am using Cinema Craft SP3, but i have a problem with VBR, it goes over 9.8mbit, even though i set the max limit at 8k. Anyone knows what to do?

[TheSkiller]

What makes you believe it shoots over the max bitrate that you set? :)
In case it is "BitRate Viewer" by "TECO", don't trust it. The bitrates it reports are wrong. You can easily tell by the average bitrate it reports - too high, and so is the max bitrate it reports.

[kolak]

Okay:)

No, i am using Cinema Craft SP3, but i have a problem with VBR, it goes over 9.8mbit, even though i set the max limit at 8k. Anyone knows what to do?

It does not, just don't trust bitrate viewers, as they are not very accurate :)

[manolito]

Mux your result with MuxMan. The MuxMan log will tell you the min, max and average bitrates quite accurately.


Cheers
manolito

[zerowalker]

i use TMPGenc, and it also tells me "Warning over 10xxx or something". And i burned anyway, and i though, hmm it works!

But at the end, it failed, you know like, Freeze and Digital artifacts, then i reseted.

It didn´t play the last 30 secs, and it probably can fail in the middle aswell, just that it didn´t happen the first time as it was, buffered enough or something.

[kolak]

SP3 does not go above specified max bitrate. Your validation methods are poor and inaccurate at all :)

[zerowalker]

Is unable to play the DVD properly inaccurate;S?
how do you suggest i make a proper accurate method then;D?

[Mug Funky]

if a disc doesn't play, a disc doesn't play.

CCE will fail a video that it can't encode within spec (it's ridiculous that it happens at all, usually on sharp detailed mostly-white line drawing stuff, but that's with SP2).

TMPGenc's muxing is probably to blame here - it's made a fool of the bitstream.

either that or your audio bitrate is too high or you're stuffing too many streams in there.

video (one angle) plus all audio cannot exceed 10.08mbps. you'd be mad to let it above 9.8.

with multiangle (per angle), it drops to total of 8.0mbps including all audio and subtitles.

[kolak]

Is unable to play the DVD properly inaccurate;S?


It can be very easily :)

Exactly- if SP2/SP3 can't encode video keeping DVD spec/bitrate restrictions it will simply stop with an error. It happens very rarely, but I've seen it.
TMPEG uses header info (not the best muxer than), which in case of Cinemacraft encoders is always (when DVD target is chosen) set to 9,8Mbit for better buffer management (better playback). Even so real bitrate will never go above max specified (it's also explained in manual). If you have playback issues than don't use high bitrate, lower burning speed, use better DVD-Rs or use different muxer. It can be also player specific problem. I use Scenarist and done more than 3000 DVDs and never had a single playback issue with streams from Cinemacraft encoders. Many of these discs went through Testronic lab and were tested on 25 most buggy players.

[*.mp4 guy]

Max gop length for NTSC with pulldown (23.976 fps) is 14 frames, not 12. 14 / 23.976 = ~0.58, which is less then 0.606 and therefore legal. I don't know why the defaults in cce are so wonky, but you really need to keep a close eye on what its doing to get it to perform well.

[kolak]

Yes, but than you may end up with "strange" GOP structure. With SP3 actual GOP structure changes anyway, so not sure about reasons for those default settings.