3D SMPTE and Subtitle Z-Depth/Parallax

[Guddu]

I am preparing a 3D SMPTE DCP. The subtitles have been authored normally and, as far as I understand, they are currently displayed at the screen plane (Z = 0). I generated a VF and did a test screening today.

While reviewing the film, I noticed that in some scenes certain objects extend in front of the screen causing the subtitles to appear behind those objects.

I have a few questions:

Is it generally recommended to leave all subtitles at the screen plane, even if some foreground objects occasionally appear in front of them?

Would it be better to move all subtitles slightly forward (negative parallax) so they never appear behind foreground objects, or would that introduce unnecessary eye strain and viewing discomfort?

Does DCP-o-matic support assigning different depths to subtitles on a scene-by-scene basis, or is this something that requires some other subtitle authoring software?

For those of you who regularly master theatrical 3D DCPs, what is considered current best practice? Do you typically:
Keep all subtitles at screen depth OR Apply a small forward offset globally OR manually adjust subtitle depth only for problematic scenes?

My goal is to produce a release that is as comfortable and professional as possible while remaining compatible with standard cinema projection systems.

I'd appreciate hearing about your real-world experience and recommendations.

[Guddu]

I have a follow-up question regarding this.

Suppose I generate a SMPTE XML subtitle file externally that already contains the required stereoscopic information (for example, a Z-position or stereo positioning value for each subtitle cue).

If I import that subtitle XML into DCP-o-matic and create a 3D DCP, will DCP-o-matic preserve that information and include it unchanged in the resulting DCP, or will it ignore or overwrite the stereoscopic positioning and render all subtitles at its default depth?

In other words, if the subtitle XML already contains per-subtitle depth information, is DCP-o-matic capable of packaging it transparently, or does it regenerate the subtitle track in a way that loses those values?

[Wolfgang Woehl]

1st-hand experience here, if somewhat limited. At Filmmuseum München we produced a couple of reconstructions/restorations of stereoscopic works from the 40's and 50's. For subtitled versions we do extensive positioning work to avoid Z-space conflicts. In all cases X and Y are involved, as well.

Yes, subtitles should not sit "behind" scene objects - super off-putting when they do. Also, usually, there is no "safe" Z-value for subtitles. If you want comfortable and "good", then there is no way around specific positioning per subtitle.

There are many aspects to this. Two random picks:

In a given image, the Z-space distance between where the viewer's gaze is lead to by the stereographer and a subtitle, should be small. If it is large, the viewer has to "look" at two different things. A viewer's gaze is not automatically lead to the max negative-Z object, could be somewhere way "deeper" in space and you'd want the subtitle "close" by. Which may, in turn, involve X-Y positioning.

A lot will depend on the stereography of the work, its camera movements and visual activity. This subtitle positioning can turn into delicate choreography.

Couple of practical things:

Having ZPosition etc. in ST 428-7:2014 is one thing - vendor's render engines are another. I burn in.

These days we're using Resolve and a polarizing monitor off a Decklink. We add rendered subtitle images, virtually convert to stereoscopic assets and can then go through all the subtitles for positioning - it is a super-nice workflow. You could use a different 3D display mode on Resolve's color page, e.g. anaglyph, and make that work if you don't have a 3D monitor.

Your follow-up suggests that you may have solved the positioning issue. Don't know about DCP-o-matic's handling of ZPosition etc. I burn in

[Guddu]

Thanks Wolfgang for your response. I am going to do the positioning subtitle by subtitle. I was reading online and at the moment, using Anaglyph and Red/Cyan glasses are going to be my only aid.

I have not solved or figured out the positioning issue entirely yet but I came across this tutorial at https://youtu.be/lMQRn0h_Efw?si=5Eg6LlJDjn1U4fad and I feel that this is very close to the "nice" workflow that you mentioned in your message.

So in the end I might be able to export the SMPTE Subtitle XML file with proper Z-position or stereo positioning value for each subtitle cue but the one thing that remains to be confirmed is if I create a 3D DCP, will DCP-o-matic preserve that information and include it unchanged in the resulting DCP, or will it ignore or overwrite the stereoscopic positioning and regenerate the subtitle track in a way that loses those values and render all subtitles at its default depth?

Also you mentioned that having ZPosition etc. in ST 428-7:2014 is one thing - vendor's render engines are another. You generally opt for burn in. What is the input that you generally give for the subtitles? A SMPTE XML file or a Transparent Video file with subtitles burnt in?

[Wolfgang Woehl]

Wrt "burned-in" subtitles: ja, while not impressive on the efficiency front, burned-in subtitles never fail. Ever. I like that.

[Guddu]

Did another test and I can confirm that Zposition is duly passed on to the final result.