XVID 23.97 FPS -> DVD conversion

[alexVS]

I have a XVID movie (.avi XVID+MP3) that has 23.97 FPS.
What is the best way to convert it to DVD and have the best quality? I use Canopus pro coder 2. So what's the choise should be?
1. Convert to PAL 25FPS
2. Convert to NTSC 29.97 FPS (here's the problem cause audio synchronization fails)
3. Convert to NTSC (23.97->29.97 pulldown) (also the problem: I receve some jerky video)

May be something else? What's your advice?

[foxyshadis]

You're not doing 3. right if you get jerkiness, probably doing a progressive pulldown instead of interlaced pulldown. You could use dgpulldown to do it after the 24fps encode anyway; procoder undoubtedly has support for flagged pulldown but dgpulldown is very good at it.

For 1. you'd have to re-encode the audio, but I guess you'll have to anyway.

NTSC vs PAL depends entirely on the region, but if you have a multiregion player & tv all that matters is one that works. 3 would be quickest but 1 won't take that long either.

[Awatef]

I would go for 3, since it retains the original cinema framerate.
2 is unconceivable at all.
1 only if it has to work on any PAL TV (not every PAL TV supports NTSC, that's for sure). Here you have to accelerate the audio. The best tool being Prosoniq Time Factory, followed by WaveLab. Those two can make the voices sound the same as the source thanks to pitch correction, without distortions.
Don't you dare convert to 25fps with blending! That will ruin the video quality! Just accelerate the video to 25fps!

[alexVS]

Don't you dare convert to 25fps with blending! That will ruin the video quality!Here I want to understand. Canopus recognise the source as non-interlaced. I guess I have to trust him. (Can XViD be interlaced at all?). And the output is always upper field first (like on every DVD). So where can I make mistake?

[Mug Funky]

@ Awatef:

that sounds like a challenge :)

for a lossy source such as an xvid (probably with reduced resolution), blending up to 50i is quite acceptible so long as the mpeg-2 bitrate is high enough and it's encoded as interlaced (ie alternate-scan, interlaced motion search, interlaced DCT, etc. procoder gives a few options for it).

option 3 is the nicest choice - straight resize of the video, then soft pulldown with DGpulldown or plain old pulldown.exe. more encoding efficiency than PAL, no need to stretch audio, no desync, etc.

[Awatef]

Alex, put your video in VirtualDubMod first, extract the audio to a separate file, go to VIDEO-FRAMERATE and input "25" in CHANGE TO and save to a new (audio-less) file with DIRECT STREAM COPY enabled.
You treat your audio in WaveLab for example and accelerate the audio by 4.27% to make the file as short as the 25fps AVI.
Then load the 25fps AVI and the new audio file in Canopus or whatever MPEG-2 encoder you're using and encode to PAL progressive (CCE may be the better choice for MPEG-2 video encoding btw)
The 50i way as explained by mug is a big NO NO. As long as you can keep the video progressive, never transform it to interlaced!

But, as foxy and mug said, 3 would be best if your TV supports NTSC, and even better, supports NTSC progressive.

[Mug Funky]

Can XViD be interlaced at all?

yes it can. it's not optimal, but it's a lot more efficient than mpeg-2 at the same bitrate :)

with mpeg-2 there's not much difference at all between interlaced and progressive.

The 50i way as explained by mug is a big NO NO. As long as you can keep the video progressive, never transform it to interlaced!

with all due respect, that's quite a short-sighted view. there's a lot of reasons one would want to keep the duration the same as the NTSC original. of course a progressive encode is preferable, but not always possible, and on a CRT they look no different at all (if done right :)). remember most people out there (including me) still have CRTs, and progressive screens are required to deal with interlace anyway (buyer beware though...).

now in alexVS's case the obvious choice is soft-pulldowned NTSC, so the PAL 50i thing is just academic.

[vr3m]

Canopus is a new name for me, and i'm new in this world,but did you thing to use DivX2Dvd?

[Awatef]

Hmm... I don't know about that "lots of reasons". The only reason I can think of is dialogs being a little bit faster. I would say there are far more reasons to keep the signal progressive especially since progressive scan capable DVD players and HDTVs are big in coming.
Even standard TVs will show a crisper video with 25p than with 50i, 25p will have better motion too.
Add to that that computer monitors don't like interlaced video at all.
As far as bitrate goes, and if my guess is right, he'll be encoding on a regular DVD-5, so we're expecting middle-range bitrates only, so I hope for him that 50i remains academic :)

[Mug Funky]

Even standard TVs will show a crisper video with 25p than with 50i, 25p will have better motion too.

well i wont get into the subjective issues around "better motion", but if you mean "half as much", then you are right. as far as crispness goes, that's a spatial thing. interlace v progressive is a temporal thing (at least on a CRT that scans interlaced anyway). an interlaced picture can be just as crisp, and in fact an IVTC + blend to 50i will look nearly identical on a CRT. i've tested this many times.

computer screens are progressive, yes, but decent DVD player software can handle it - it's part of the spec after all, so software that can't hack it shouldn't rightfully be called "DVD" software at all.

as for reasons to maintain the speed of the original:

- audio doesn't need changing
- subtitles will not need re-timing (this can be a pain, though it sounds quite simple. when edits are involved it can be a nightmare and a big time-waster)
- a realtime rig cannot change the speed regardless unless there's equipment that can be undercranked (ie a PAL deck running at 48i but recording onto 50i, and a standards-converter with a clever IVTC, or just HD equipment all through).
- client might not like the idea of the running time changing (usually they're smarter than that, but they have the final call).
- the source could be a mix of film and video, and the video might have priority (besides which it's exceedingly difficult to distinguish between film and video reliably, as i'm finding out...)

besides which, even on a DVD-5, unless you're talking about LotR or kenneth branagh's Hamlet, then there'll be enough bitrate.

[shirohamada]

i think dgpulldown can soft pulldown to 25fps.

[alexVS]

Turning back to what was said. So XviD can be interlaced. I didn't know it before. But how to find out if it's interlaced or not?

For MPEG2 there is DVD2AVI or DGIndex. It shows if source MPEG2 interlaced or not.

What about XviD? Where can I determine if it's interlaced or not? And what field goes first (upper, bottom)?

[Mug Funky]

dgindex et al just reads what's written in the stream. so the only way to know for sure is to look at the video.

however, if it's an xvid at 23.976, you can be pretty sure it's progressive.

as for mpeg-2, visual inspection is the only way to go - big budget movies aside, most DVDs are encoded as interlaced just to be on the safe side, even if what's on them is progressive or telecined (it's far worse to encode interlaced as progressive than to encode progressive as interlaced. the latter will have slightly fuzzier chroma, where the former will have very poor quality and chroma will be munged completely. as such most of the time the person encoding just leaves it on interlaced mode).

it only takes a few seconds to open something up and scan through it for interlacey-looking frames, and a few seconds longer to spot a 3:2 pattern (you'll see 3 progressive frames to 2 interlaced usually) meaning you can IVTC and encode at 23.976fps

good luck :)

[Awatef]

Oh well mug, as you have to pay big bucks for a nice car, you have to spend big time for great video quality. If he's not ready to spend enough time for progressive PAL, he can use the 50i solution for all I care, but I will never do it, since I hate blending as much as I hate interlacing, especially if it's avoidable.
And no matter how good the DVD player may be, it will never deinterlace it perfectly, since blending is a big bastard (welcome motion jerkiness).
You can barely get away with it if watched on a normal TV, since 24p to 50i is much more "motion-friendly" than telecined 30i to 50i, but sorry, I still get it that there is something wrong with the video: blending stands out directly and the motion is not regular, especially on pan scenes.
By the way, all PAL movies I ever watched are encoded using the speed-up method, so it's pretty accepted, since it leaves the video quality/crispness intact (I insist on using the word "crispness" since 50i will have a blurrier picture compared to 25p). I can see myself moaning if I get an interlaced PAL movie! (Curse all Anime companies that produce interlaced PAL DVDs even if the source is 100% 24fps)

[Mug Funky]

Curse all Anime companies that produce interlaced PAL DVDs even if the source is 100% 24fps

you realise that's me you're cursing (though i just checked and africa isn't region 4).

the only 100% film stuff you'll find in english would be movies, and i'll speed those up if i can. anime series are always a mixed bucket. try to ivtc and an AV tester will inform you that there's bad scenes. i have a rather long avisynth script that handles it pretty well, and it has the option to output progressive (where possible) or interlaced. i'll choose interlaced most of the time because it means the duration doesn't change. that means audio and subs don't need retiming, and a host of other advantages.

if these shows were mastered with PAL in mind, i could speed up everything, but it's just not like that in japan, or in fact any NTSC country (look at the US - you'd be hard pressed to find anything at all that's PAL capable, so why would an editor waste time on it?).

however, i do actually watch my encodes, and my opinions are based on that - IVTC-to-50i is almost indistinguishable from 25p but for the speed difference (if they're done properly).

just please remember that PAL anime companies generally do their best - the people that make the discs are fans of the shows they produce, and no-one here would wantonly sacrifice quality.

[Awatef]

The best way remains to go back to the original 16mm or 35mm source for companies anyway, because there are many shows that are 24fps all the way through (especially older ones)
In case there are no such sources (difficult to imagine?! at least digital 24fps sources?!), many shows still are ivtc-able to at least 99%. I can definitely live with 1% blending instead of 100% ;)
Of course if the show is a mixed bag (Gundam Seed for instance), you can't really blame any PAL Anime company for producing interlaced blended stuff...

Anyway, in the case of Alex, we're talking about pure 24fps. The only time consuming part is the audio speed-up, and I would gladly do it, since pitch correction works like a charm. The video part is a joke, it's just a special copy/paste. In case there are subs, the speed-up is a question of seconds, and I know what I'm talking about ;)

I can upload some DBZ 25fps progressive stuff with subs if you want...

[Paulcat]

If you are starting with a 23.976 avi file, personally what I'd do is this:

1) Extract the audio as an mp3 (Avi-Mux GUI works)
2) Use Nero toolkit or BeLight to convert to a stereo wav file
3) Use BeLight again to convert the wav to a stereo ac3 file

Converting directly to ac3 with BeLight causes a big drop in volume, the two steps does as well but not as much.

4) Convert your avi (video only if your encoder can), leave it at it's current framerate, and encode it as progressive (non-interlaced) in either PAL or NTSC, up to you.
5) Use DGPulldown to alter the framerate of the .m2v file from 23.976 to 25 (for Pal) or to 29.97 (for NTSC)
6) Use your DVD authoring software to complete your project (MuxMan works and it's free)

Paul

One of the reasons for the jerky playback is when your encoder converts from 23.976 to 29.97 it may just duplicate every 5th frame to fill in the spaces, and our eyes can pick that out. TMPGEnc does that, thats why I encode at the original framerate and use DGPulldown to fix it after.

Also with this method, there is no dicking around with the pitch and duration of the audio, as long as they have the same running time, you're all set.