Feature Film - Croping and sound sync

[Huiuuu]

First of all , thank you for this wonderful program and active from what it seems board.
My first post here just to save some time from experimenting my self.

I am in Hurry to make a Dcp for a feature film . It needs to be at 24fps 1.85 flat with 5.1 .
I have a master file @ 25fps at 16:9 with 1.85 flat letterbox.
If i am right , to make the Dcp at 1.85 flat , i will need to crop 21 from top and 21 from bottom , and scale to: set to No stretch. ( I am worrying if the movie gets stretched ) it seems to work right...
To make the DCP at 24fps , i will not mess with the content but i have to set the DCP settings to 24fps , am i right?

Now with the sound. I have from my sound engineer the 6 files of the 6 channels as discreet mono channels. After i import them i set the channels in the Content-audio tab. The sound is at 24fps, so already stretched. And here is my biggest problem , if the sound is already at 24fps , the program re-strech it when i change the settings in the DCP tabs to 24fps. Resulting in unsync sound. Whats the right way to fix that? I dont want to use the 25fps stems because i trust more my engineer on streching the audio. Does changing the audio fps at Content-audio tab to 24fps will solve the problem or it will result in double restreching ?
I would love to find a way so the program will only strech the video , leaving the audio intact.

And one last fast question. The projectionist asked for 24fps DCP (the movie is at 25fps , and i hate my life for slowing it down) , that means most likely that the projector is interlop standard. Should i make the DCP interlop or , 24fps Smpte works fine in interlop projectors?

Thanks in advance ! I hope i didnt tired you.

[Carsten]

Hello - I guess 16:9 flat letterboxed means a 1920/1038 image letterboxed in a 1920/1080 frame? In that case, crop 21pix from top and bottom, and scale to flat, choose the flat container. If you are concerned about the scaling - it makes little sense to create a flat feature, but then put it into a 1080p frame for cinema presentation - it will end up with black bars both top bottom and left right - not good. You need to fill a 1998/1080 frame with your 1920/1038 content. You can't avoid scaling here.

Yes, changing the frame rate reference of all 6 audio files to 24fps should leave the audio untouched. Check back the frame rate under DCP tab - sometimes it seems DCP-o-matic resets the DCP frame rate to 25fps when you re-reference a source file. When you open timeline view, you should be able to see the length of video and audio matching. Also, select all six audio files one by one and check under audio if the bottom shows 'Audio will not be resampled'.
In the future, you do not need to order stretched audio - there have been comparisons of common resampler performance, and the one used by DCP-o-matic is among the best. Such a small adaption of sample rates is usually not a big thing anyway.

What is your master video format - image sequence, or a video container/codec?

You should create an interop DCP. 25fps works fine in interop, too, but there are some older servers around that can not play 25 fps, not in SMPTE, not in interop. This projectionist may just play safe, but it may well be that their server actually lacks 25fps capability.

In the future, I would suggest, if you have a 25fps master, you add 25fps video and 25fps referenced (unstretched) audio, then create both a 25fps and a 24fps version by simply changing the frame rate under the DCP tab. That takes care of everything. Then you are free to choose which DCP to send/play.


- Carsten

[Huiuuu]

Thank you VERY much for the fast reply!
Seems that i did everything right, i hope it will test out well!
Audio will not be resampled and did the right thing with the ratio, i knew it will be scaled..
My master video is a DnXhd at 185 kbps on .Mxf does it matter ?
Should the next time export the Master video file to a flat ratio from the NLE program so the program doesnt need scaling?
Is there anything else i can do improve the dcp workflow ? Like image sequence or other codec ?

Thank you again for your fast answer !

[Carsten]

DNxHD is fine. The question is, does you camera capture at max 1920 width, or can it do 1998/2048?

While it is 'okay' to put 1920/1080 images pillarboxed into a 1998/1080 container, it is not advisable as a general workflow, as you will see the black pillarbox bars in most cinemas. So, while it is equally acknowledged that scaling should be avoided, at least if unnecessary, it is still better to crop and scale 1920/1080 footage to a full flat DCI frame (1998/1080). Scaling algorithms in general do a nice job nowadays (I think DCP-o-matic still uses bicubic).

It is, of course, up to you - if you feel that image quality is the most important aspect for you, and consider your content highly delicate in resolution and structure, you may feel that the pillarboxing is less critical than scaling, and try a 1:1 pixel mapping using either 16:9 (in the case of 1080p), or no stretch/no scale.

Of course, if you have access to a cinema for testing, you can easily try it both ways and see wether your concerns are overridden by your own viewing experience. When I create 1080p content for our own cinema, I usually do not scale and leave the 1920 width untouched - because I can create custom presets on your projector and masking, and so get rid of the pillarboxing by fine-adjusting our screen masking. When I create a flat or scope DCP to be presented in other cinemas, I usually scale 1920 width to either flat/1998 or scope/2048, so the DCP is 'universally' compatible with these two common formats.

Another tip - I usually avoid having separate mono-files for audio, because it will cause more effort to adjust settings in DCP-o-matic, as every setting has to be copied to every single file. It's easy to introduce errors that way (e.g. miss the 24fps reference for just one file). If the channel assignment is obvious/properly documented, then multichannel files only need a single trimming, gain, frame rate reference, etc. value and there is less risc for error.
Also, with a multichannel file, a proper audio graph including all channels (incl. LUFS) can be computed under the content/audio tab, not just under the DCP/Audio tab. As LUFS has become an important measure for proper audio loudness, having it available as early as possible is a nice-to-have.
MXF is perfectly okay, as a matter of fact, you can easily create a 1:1 multichannel version of your audio by simply adding all audio files to the project, taking care they are set up in parallel, then create a DCP (without any image content added). This will quickly create a (very small) DCP with a black video MXF, and a multichannel audio MXF. You can re-import that audio MXF into DCP-o-matic, and use it as the project audio together with the video.

- Carsten

[Huiuuu]

Seperate mono files , was a pain in the ass all along, as i couldnt easily create a 5.1 H264 or other 5.1 delivery formats easily...
I would definitely ask for a multichannel audio the next time, even if they want to accompany them with seperate channels as a bonus...

If the channel assignment is obvious/properly documented, then multichannel files only need a single trimming, gain, frame rate reference

What you said raised me a question , when do you need to trim or change gain of an audio? Is this happening in cases ? If the Masters are done right , wouldnt that be enough ? I didnt touch those fields in my case cause timings between audio and video are right as it is and i thing that the sound guy did the mix with dolby standards in mind. Or i missed something?

I could stay here asking questions whole day , thanks again for answering i owe you one

[Carsten]

It clearly depends on the capabilities of the audio mastering facility, and what you ordered from them. If you ordered them to make a cinema compliant mix, they may have applied reference levels or LEQ(m), and you should not touch the gain settings. If, after a test screening, you feel that the levels are way off, you may need to adjust them in a second run (if you kept the project complete, DCP-o-matic will only recompute the audio and re-use the encoded video).

Sometimes source content is mastered to 0dB FS, or includes bars&tones, 2pop, whatever, so you may be forced to adjust levels or trim out portions.

I would be interested to learn the LUFS value of your feature - if you don't mind, re-open the project, check in prefs wether 'find integrated loudness...' is checked, go to DCP-Audio, and click 'Show Audio' - give us the 'Integrated Loudness' (LUFS) number.

- Carsten

[Huiuuu]

well it needs 5 more hours to finish the export, can i do that while dcpomatic is transcoding , or i have to w8 for it to finish?

[Carsten]

Let it first encode to end. Depending on your pref setting, that analysis may have been done already, and may give results within a second. However, it you first need to enable it, a full new analysis of a 1.5 hrs 6 channel piece could take quite a time and slow down the encoding as well.

- Carsten

[Huiuuu]

Ok i now audio is analysed i will come back with the results.

I played back the dcp on dcpomatic player , and its doing some stutters every once and a while , and drops a small amount of frames even in quarter decode. That means that the dcp is problematic and will do the same at the cinema?

[Huiuuu]

So Sample peak is 0.43dB
true peak is -0.40 db
Integrated loudness -25.41 LUFS
Loudness range 15.16 LU