Convert 24frames to NTSC and PAL

[PMedia]

Could someone in the know PLEASE give me the final how too on conversion. I've read so many thread that all seam to differ.

I have a 24fps film 24bit at 48K I need to convert a PAL version 25fps and a NTSC version 29.9fps. I'm using Nuendo 5

[danijel]

OK, I just went through this horror again this week (only in the opposite direction 25->24), so I'll tell you how it went.
This was my first time-stretching using MPEX 4 built into Nuendo 5 (poly-complex algo). I thought I won't have the time to do separate passes for every stem, so I took the whole mix, reel by reel, and slowed it down to 104.16666% of the original duration. At first I thought it sucked, so I went back to my trusty old Nuendo 3 (which uses MPEX 2), but it turned out that some of the music cues in this film were actually very prone to time-stretching problems, and the result was maybe a bit better, and maybe not at all. The color-grading department had some problems and I found out I have 2 more days. OK, I took the four stems (DX, FX, BG, MX), slowed each one down separately, and everything was perfect except for the aforementioned music cues, which sounded much better when dealt with in a separate pass, but still, there were noticeable bumps and glitches. So, as I had some time for experimenting, I decided to try out every major time-stretching algorithm out there. This page helps a lot, even if you don't understand French:
De l’étirement et de la hauteur
To cut the long story short, it turned out that those cues which are pad-based, or drone-based (as opposed to rhythmic) were better off when I first resample them to the correct length (without pitch-correction), and then put through a real-time pitch correcting plugin, like zplane's Elastique Pitch (thanks to Steven Ghouti for the tip - he has a similar workflow, with outboard pitch correction). The rhythmic cues, however, did much better with MPEX4 in my case. As for all other stems, there was not one issue with MPEX4.

As for the workflow (your main question) - whether you're stretching stems or the complete mix, first make it a poly-wav - if you stretch the separate mono wav's, you'll have issues with stereo downmixing because the inter-channel phase will be screwed.
So, put the multichannel wav on the timeline, rightclick -> Process -> Timestretch. If you're going from 24->25, enter 96% (that's 24 divided by 25), check 'Correct Pitch' (or something like that, I'm not at Nuendo right now), choose 'Poly Complex' algo, and off you go. In the opposite direction, enter 104.166666 (that's 25/24).
For 24->23.97 I don't correct pitch, I just resample the file to 48048 samples per second, the composer will not notce the pitch change

One more thing, if you time-stretch the whole thing, with the two-pops, the time between the leading two-pop and the FFOA, and between the LFOA and the ending two-pop will stretch too, naturally. The problem is, sometimes in PAL, we use 49 black frames + two pop, and sometimes it's 47 + two-pop, so get in touch with the picture department and make sure you're in sync with them.

[PMedia]

Originally Posted by danijel

OK, I just went through this horror again this week (only in the opposite direction 25->24), so I'll tell you how it went.
This was my first time-stretching using MPEX 4 built into Nuendo 5 (poly-complex algo). I thought I won't have the time to do separate passes for every stem, so I took the whole mix, reel by reel, and slowed it down to 104.16666% of the original duration. At first I thought it sucked, so I went back to my trusty old Nuendo 3 (which uses MPEX 2), but it turned out that some of the music cues in this film were actually very prone to time-stretching problems, and the result was maybe a bit better, and maybe not at all. The color-grading department had some problems and I found out I have 2 more days. OK, I took the four stems (DX, FX, BG, MX), slowed each one down separately, and everything was perfect except for the aforementioned music cues, which sounded much better when dealt with in a separate pass, but still, there were noticeable bumps and glitches. So, as I had some time for experimenting, I decided to try out every major time-stretching algorithm out there. This page helps a lot, even if you don't understand French:
De l’étirement et de la hauteur
To cut the long story short, it turned out that those cues which are pad-based, or drone-based (as opposed to rhythmic) were better off when I first resample them to the correct length (without pitch-correction), and then put through a real-time pitch correcting plugin, like zplane's Elastique Pitch (thanks to Steven Ghouti for the tip - he has a similar workflow, with outboard pitch correction). The rhythmic cues, however, did much better with MPEX4 in my case. As for all other stems, there was not one issue with MPEX4.

As for the workflow (your main question) - whether you're stretching stems or the complete mix, first make it a poly-wav - if you stretch the separate mono wav's, you'll have issues with stereo downmixing because the inter-channel phase will be screwed.
So, put the multichannel wav on the timeline, rightclick -> Process -> Timestretch. If you're going from 24->25, enter 96% (that's 24 divided by 25), check 'Correct Pitch' (or something like that, I'm not at Nuendo right now), choose 'Poly Complex' algo, and off you go. In the opposite direction, enter 104.166666 (that's 25/24).
For 24->23.97 I don't correct pitch, I just resample the file to 48048 samples per second, the composer will not notce the pitch change

One more thing, if you time-stretch the whole thing, with the two-pops, the time between the leading two-pop and the FFOA, and between the LFOA and the ending two-pop will stretch too, naturally. The problem is, sometimes in PAL, we use 49 black frames + two pop, and sometimes it's 47 + two-pop, so get in touch with the picture department and make sure you're in sync with them.

Hey there

So the PAL version (25 frames is perfect but the NTSC is not) For the NTSC version I converted the 24 frame original at 99.9% but going on your maths for PAL I think that is wrong.

NTSC is 29.97 is it not? if so 24 / 29.97 = 80.008%

Is this what I should be speeding it up to?

[minister]

Originally Posted by PMedia

Could someone in the know PLEASE give me the final how too on conversion. I've read so many thread that all seam to differ.

I have a 24fps film 24bit at 48K I need to convert a PAL version 25fps and a NTSC version 29.9fps. I'm using Nuendo 5

I just did this, and it can be quite easy. HOWEVER, my film assumed a frame = a frame conversion. If your film did not do this, then the following does not apply. If your film used a second = a second, OR, simply doubled frames to get to 25, the following also does not apply. You have to ask the post super or editor or transfer person.

For a frame = a frame conversion, your math is simple:

24 --> 23.98 (which is equivalent to 29.97)=-0.1% slower. So your clock source at 47952Hz (which is .1% slower than 48000)
24 --> PAL 25 is +4.166% (faster). Or, your clock source at 50000

So, you need to tell Nuendo about the above SR values for the source of the files to be pulled to a 48K SR, OR, you need to use a Time Compression/Expansion Algo to change the speed.

I recommend the SRC trick. Some TCE programs are aliasy or just plain wrong.

Method in Pro Tools (sorry, don't know Nuendo, but might apply). Select file to import (24fps mix). Chose to SRC file. For source SR, select 47.952 for the 24-->23.98 convert. It will then pull down the file against a 48k clock and your new track will be .1% longer. Similar for the 24 to 25 file. Select 50000 for the source SR and convert it to 48K. the file will be pulled up (shorter) relative to the 48k clock and your 24fps file.

[PMedia]

Sorry if this sounds silly but I live in PAL world so NTSC is a bit strange to me. I thought it ran at 30fps until I found out it was actually 29.97fps. What confusses me is when you say 23.98 is the same as 29.97fps ???? how does that work

I want to time stretch just like I did for PAL as I got really great results. So the SRC option is not what the DCP house wants.

Any suggestions

[PMedia]

May have found the answer in this thread
29.97 and 23.98 SMPTE timecode -- The definitive explanation

[Branko]

I would definitely recommend Prosonic's Time Factory 2, DIRAC algorithm. Digi's X-form is also very good, if you have enough time...

[neilwilkes]

All these solutions alter the audio though, don't they.
Artifacts, glitches, pitch changes, running time issues - seen the lot.
I even have some EMI DVD discs where some of the tracks are playing too slow at the wrong speed & pitch precisely because of this converting.
(Check out "Best of Bowie, PAL version, and watch the Bob Rock films - all slow & pitched wrong)

Best bet is to get the film properly rendered in the first place & leave the audio alone, I think. Footage can be reframed easily enough - we took a 1920x1080/25p film to 24p and did not change the running time by as much as a millisecond from the original, completely artifact free, and once you got 24p you got enough for NTSC DVD as you can author with the 24p footage.

It all depends on what you have to do, what it's for, what the budget & time constraints are etc.

[Geert van den Berg]

Originally Posted by neilwilkes

Best bet is to get the film properly rendered in the first place & leave the audio alone, I think. Footage can be reframed easily enough - we took a 1920x1080/25p film to 24p and did not change the running time by as much as a millisecond from the original, completely artifact free, and once you got 24p you got enough for NTSC DVD as you can author with the 24p footage.

What software solution did you use for this reframing? And how often have you done this?

In my own experience changing the video frame rate without changing its running time works okay on simple material but material with lots of detail and movement (and pans) doesn't look so well when altered (easy to spot artefacts of the frameblending). It's done, but I don't think it's more common than altering the audio for a good reason.

In the case of video and audio. IMHO for a feature, video is king, so in most cases it's often easier to change the audio. Unless it's a video of a music performance, then it's another thing. Then a different pitch is more annoying.

Usually I also use the SRC method nowadays, basically because I can't stand the artifacts of the pitchshift algo's. Tried quite a lot of different plugins, thought the Waves SoundShifter was quite good, but later on, I didn't like it anymore. Still too much artifacts, losing transients and turning the sound lo-res (mp3-ish).

[Branko]

Originally Posted by Geert van den Berg

...changing the video frame rate without changing its running time...

How do you do this? The running time of the film is defined by the:
a.) amount of frames to play and
b.) playback speed
which can be expressed by a simple formula:

running time = total number of frames / playback speed

How do you change the frame rate (playback speed) without changing the running time?

[Geert van den Berg]

Originally Posted by Branko

How do you do this? The running time of the film is defined by the:
a.) amount of frames to play and
b.) playback speed
which can be expressed by a simple formula:

running time = total number of frames / playback speed

How do you change the frame rate (playback speed) without changing the running time?

It's done by frameblending and/or motion analysis. The exact process I don't know. You can do it with for example Apple Compressor and I think Telestream Episode Pro will also do it. It is possible, but like I said, I never saw it without visible artefacts. That's also why vote against it, unless Neil has a solution that always works (could also be that 25 to 24 is easier and less prone of artifacts because it's a frame less per second, I only checked the other direction). And maybe a more expensive software package has a better algorithm.

[PMedia]

If I could move back to my question. The time stretch for PAL worked well. I was able to stretch each stem atmos music dialogue separately then master off a new 25 fps file - Worked great sounded great. No artifacts no pitch issues.

My NTSC version sounded great too but sync drifted more and more in the film. I used 99.9% but I'm thinking I went the wrong way and should of made it slower instead of faster. So maybe should of been 100.1%

Any thoughts?

[minister]

Originally Posted by PMedia

I want to time stretch just like I did for PAL as I got really great results. So the SRC option is not what the DCP house wants.

Any suggestions

With the SRC method, you end up at 48k, so why is that not what the DCP house wants?

Originally Posted by PMedia

My NTSC version sounded great too but sync drifted more and more in the film. I used 99.9% but I'm thinking I went the wrong way and should of made it slower instead of faster. So maybe should of been 100.1%

Any thoughts?

Yeah. As I said in my post, it is slower.


NTSC is 29.97, but for all practical purposes, 23.976 and 29.97 are the same "speed".

[danijel]

Originally Posted by PMedia

My NTSC version sounded great too but sync drifted more and more in the film. I used 99.9% but I'm thinking I went the wrong way and should of made it slower instead of faster. So maybe should of been 100.1%

Any thoughts?

Yes, it should be slower. And you really don't need to correct the pitch or do the stem by stem routine in this case, the difference is unnoticeable - just resample the whole mix at once.

[neilwilkes]

Originally Posted by Geert van den Berg

What software solution did you use for this reframing? And how often have you done this?

In my own experience changing the video frame rate without changing its running time works okay on simple material but material with lots of detail and movement (and pans) doesn't look so well when altered (easy to spot artefacts of the frameblending). It's done, but I don't think it's more common than altering the audio for a good reason.

In the case of video and audio. IMHO for a feature, video is king, so in most cases it's often easier to change the audio. Unless it's a video of a music performance, then it's another thing. Then a different pitch is more annoying.

Usually I also use the SRC method nowadays, basically because I can't stand the artifacts of the pitchshift algo's. Tried quite a lot of different plugins, thought the Waves SoundShifter was quite good, but later on, I didn't like it anymore. Still too much artifacts, losing transients and turning the sound lo-res (mp3-ish).

I did not do it personally, but we had it done by someone I trust implicitly and who has handled all our video conversion needs for 10 years now - the guy is so good it's scary, and he has forgotten more than I know.
The last one he did for us (the 1920x1080/25 to 1920x1080/24) was acclaimed as being a great looking disc too
(the first HD one we did - all the rest were SD, and never a problem)
As to what he uses, I do not know. He uses a farm of some sorts, and a very heavily customized piece of software.
I've seen him succeed where Alchemist HD failed.

I know what you mean about pitch shifting artifacts though - horrible, in the main, and the best I have found so far is the solo musical MPEX-4 one in Nuendo - forget Poly Complex in comparison as it leaves many more artifacts.
SRC is better, but once you start converting down you start to introduce aliasing problems unless you are very lucky indeed. Going up is not so problematic - it's always coming down again that is the problem.

Admittedly, our main stuff is all music based, so pitch shifting is simply not an option - we have to fiddle with the video.

[PMedia]

Originally Posted by danijel

Yes, it should be slower. And you really don't need to correct the pitch or do the stem by stem routine in this case, the difference is unnoticeable - just resample the whole mix at once.

Hey Dan

Would you mind stepping me through the SRC process in Nuendo just so I get it right first time

Be awesome if you had a minute Cheers

[danijel]

Sure, import your finished 24 fps mix as a multichannel file, put it on a new track, select it, go to Audio->Process->Resample, put 48048 in the 'New sample rate' field, hit 'Process'. It should be 0.1% longer after that. Check with the video post people if the duration matches, and export it as a new file.

You can probably do the operation on separate mono clips (for L, R, C etc) - it shouldn't screw the inter-channel phase in this case, but it's safer to do it on one file.

Hope that helps

[musikwerks]

Here's an idea...

Hire yourself a Panasonic D5-HD deck, the 3700 model with all the TC conversion options. It can run in all SD and HD rates.

Master the film with finished audio at 24fps to the D5, then do a downconvert to NTSC spec. The machine will do everything for you.

For PAL, do the same but then re-stripe your master with audio sped up by 4.1% as mentioned earlier. With or without the pitch correction, depends what the client wants. I've seen it done both ways. Often the US release will have differently pitch audio than the Aus version.

We did/do it this way for worldwide masters, for Disney, and they're not easy to please.

Not sure why people are talking about sample rates? Nothing to do with it... it should stay 48khz if you're mastering to video. It's the speed (and pitch) not the sample rate.

[danijel]

Originally Posted by musikwerks

Not sure why people are talking about sample rates? Nothing to do with it... it should stay 48khz if you're mastering to video. It's the speed (and pitch) not the sample rate.

When you resample, you get different number of samples for the same piece of audio. Now, when you play it back at 48kHz, its speed/duration has effectively changed.

[musikwerks]

Nobody I know resamples... not for any of the official masters we have done. It's always sped up by 4.1% then, if requested, pitch corrected. Usually not pitch corrected though.

Very often mixes with stems come on a DA-88/98 tape. It's quite easy to speed that up, lock it to tape and re-stripe.

[ErikG]

Originally Posted by musikwerks

Nobody I know resamples... .

That's just an illusion...
Speeding up down in the digital domain is the same as SRC either you resample through the digital I/O (requires hardware with good SR Converters) or you play over analog and rerecords through two D/A A/D conversions.
The difference is that either the hardware or the software handles the sample interpolation.

On the old (awful, hated them) DA88's the digital out would be simply speeded up +4.1%

[musikwerks]

And that's exactly what I said we do... speed up the DA88's. No major feature we have had has sent re-sampled (as per the description of re-sampling earlier) soundtracks. And these have gone to Disney for worldwide release with smiles from Disney.

[danijel]

Originally Posted by musikwerks

And that's exactly what I said we do... speed up the DA88's. No major feature we have had has sent re-sampled (as per the description of re-sampling earlier) soundtracks. And these have gone to Disney for worldwide release with smiles from Disney.

Well, there are many ways to skin a cat. So, you don't resample in software, but in hardware. The fact that Disney doesn't mind you doing it through DA88 doesn't make a case to recommend someone to rent a bulky piece of equipment to do something that is built into every DAW.

BTW, what does re-stripe mean?

[josh broome]

Re-stripe is another way of saying re-record, or replace, the existing audio. If someone hands you a tape and says "can you stripe this in?", it would mean to record the audio off the tape into a DAW.

[danijel]

Originally Posted by josh broome

Re-stripe is another way of saying re-record, or replace, the existing audio. If someone hands you a tape and says "can you stripe this in?", it would mean to record the audio off the tape into a DAW.

Thanks. Does re-stripe involve modification (like sweetening), or just straight copying?

[josh broome]

I'm really not sure, it's just lingo, If I were to record my mix onto a master tape I would call that a "layback". Re-stripe to me sounds like the audio is the same or changed slightly. If I were to simply "stripe" a tape I would put black, silence and timecode on it, some people call that a crystal, some call it a basic... I never heard of "faxing" until the other day, when you're in a studio production and you check to make sure the audio signals are correctly routed to the appropriate channels on the VTR's It's called "faxing" I'm sure it's called something different somewhere by the way thanks for your informative posts, I've picked up quite a bit from you.

-Josh

[minister]

Originally Posted by musikwerks

Snip.

For PAL, do the same but then re-stripe your master with audio sped up by 4.1% as mentioned earlier.

Snip.

Not sure why people are talking about sample rates? Nothing to do with it... it should stay 48khz if you're mastering to video. It's the speed (and pitch) not the sample rate.

How do you think the Panasonic Deck accomplishes this task and keeps the audio at 48k? There must be a clock.

What do you think governs the speed of video? Frames, right?

What governs the speed of audio?

You may not see the process of the SR being clocked differently, but it is "under the hood".

You can use a Time Compression/Expansion plugin to accomplish the speed shift. I have seen widely varying results. The SRC trick has been flawless for me. And yet, I wonder how the TCE plugin changes audio speed 4.1%....could it be???......

[Takeshi]

Hi everyone,

I agree that just changing the playback speed (varying Sample Rate on digital systems) is the most accurate method, and we always do it that way when 0.1% pull down is needed. It's mathematically correct and we've never found any problems with that.
Nevertheless, we live in a PAL country (25 FPS) so we get to do a lot of 4.166% pull ups (24 to 25 FPS). In this case, Sample Rate Conversion is still the most accurate, but usually the resulting pitch shift is unacceptable. For this task, we have tested many plug ins and processors for pitch correcting and found out Prosoniq Time Factory being the best sounding and duration accurate, to our judgment. It respects phase coherency awesomely, that's very important when shifting LtRt programs.

Best,

[danijel]

Originally Posted by josh broome

by the way thanks for your informative posts, I've picked up quite a bit from you.

Glad to hear that. Reading this forum for the past couple of years has been outrageously useful for me too. Being a little off-road, without the opportunity to learn by watching others do these things, I'd feel very inconfident about my skills with only what I've picked up at the film school

[PMedia]

So just for the record what you are all suggesting is

24p to NTSC
24-29.97 Sample Rate Convert it to 48048 Hz then re-export at 48000Hz


24p - PAL
Option 1 Pitch correct using Algo's in this case Mpex Solo Musical
24-25 Using Time stretch speed up by 4% or in Nuendo 96% compressed and re-export at 48000Hz

Option 2 No pitch ie sounds higher
24-25 Sample Rate convert 46080HZ then re-export at 48000Hz