Blackmagic Forum

The subtitle feature is awesome in Resolve.

However, subtitles are rendered in 2D on the screen plane interfering with the stereoscopic movie.
Is it possible to apply a corrector node to a subtitle and change the position in space by adjusting convergence? In the export the convergence could be translated to z-depth for the subtitles in digital cinema.
I would be very glad to know if there is a solution for this workflow.

Many thanks in advance.

Same question for me.

How should this work? If you set convergence for a 2D object (i.e. have a horizontal offset for both subtitles), how should your brain determine if the text is in front or behind the projection-plane?
2D objects always live on the screen-plane since convergence only works with 3D objects.

SeldomSeenKid wrote:How should this work? If you set convergence for a 2D object (i.e. have a horizontal offset for both subtitles), how should your brain determine if the text is in front or behind the projection-plane?
2D objects always live on the screen-plane since convergence only works with 3D objects.

We talk about stereoscopic 3D, so we have t determine if the subtitle has to be in front, on or behind the projection plane to be in adéquation with the convergence of the shot.
So the goal is to set the convergence on the subtitle when you work on 3D stereoscopic films.
Of course it is useless on 2D.

Fabien Remblier wrote:We talk about stereoscopic 3D, so we have t determine if the subtitle has to be in front, on or behind the projection plane to be in adéquation with the convergence of the shot.
So the goal is to set the convergence on the subtitle when you work on 3D stereoscopic films.
Of course it is useless on 2D.

So do I. Say you want to place your 2D object (text) a given distance X in front OR behind the screen-plane. The projection offset between the two (left/right) representations of this object will be equal in both cases. As a consequence your brain will not be able to determine the correct depth.

I think you don't understand what I want to say.
Here is an example of handmade steroscopic subtitles:
https://vimeo.com/manage/videos/511662509
Check at 00:39, you will see that the placement of the subtitle is not the same on left and right eye so the subtitle is in front of the actress and does not make a visual conflict with her.
This is actually impossible to do with the subtitle tool.
We can only do it in the edit part which make it really hard to do because I have to switch between 2D estimation (in the Edite part) and 3D vision (in the color part)
It would be easier to access to subtitle convergence in the Color Part.

Fabien Remblier wrote:I think you don't understand what I want to say.

I'll try to clarify my point:
First of all, our brain uses different cues to determine depth one of them is convergence. Others are parallax, lighting and of course occlusion. Whenever these rules conflict with each other, our brain gets stressed and we don't experience the result to be pleasing. This is why only a few 3D movies (like Avatar) work and the waste majority don't.
Convergence is determined by the angle of the cameras pointing inwards. The point where the camera-beams converge is defining the projection plane (aka screen-plane). Everything between the camera and the projection-plane appears to be in front of the screen, everything behind the projection-plane appears to be behind the screen.
As a rule of thumb your subject should never be in front of the projection-plane for numerous reasons, thus your subtitles can live on the projection plane without conflicting with your subject. Only objects, which are not cropped by the screen borders can stay out of the screen without causing eye-strain.
In your example, the only reason why the titles appear in front of your subject is because they're occluding it. The brain weights the occlusion rule higher than the convergence rule. Nevertheless it cannot determine the correct depth only from convergence (which was my point).

The problem Uwe and Fabien point out is ofc real. E.g. the Avatar people did extensive (and wonderful) X-Y-Z positioning work for their in-film subtitles (Na'vi) as well as their international subtitled versions. Subtle hand and taste required.

Simply sticking subtitles to the screen plane does not work at all.

We did stereoscopic subtitling for Aleksandr Andriyevsky's stereoscopic film "Robinzon Kruzo" from 1947 - a work with many elements in negative z-space ("in front" of the screen). Due to the constraints of a very special presentation setup (no glasses required) the film has an aspect ratio of 1:0.85. Unrestricted Flora everywhere. In other words: space in this film is tight - in all directions and dimensions.

At the time we used a cheap stereoscopic monitor and a self-developed interactive tool to step through and position the subtitles. Iirc, there were two commercial solutions available which were ludicrously expensive.

I second Uwe's and Fabien's request for better integration. Nice-to-have on my list.

Wolfgang Woehl wrote:Simply sticking subtitles to the screen plane does not work at all.

Well, I did a bit of research and indeed it looks like it isn't as simple as I thought. Adjusting subtitles to appear slightly in front of the object being closest to the viewer appears to be a common practice. This somehow makes sense as the observers eyes most likely converge at the closest object.

I have to admit that I never did a 3D movie. My background is 3D computer graphics (i.e. VR). When we had to project a 2D object on a 3D screen (like a UI element) we'd always put it on the projection plane for best user convenience. This is where my suggestion came from.

Looks like 2D subtitles are a different "use-case" and I cannot generalize here. However, my very first point still holds true. Given an empty scene with only a 2D subtitle, which is not placed on the projection-plane will confuse our brain because of the behind/below dilemma.

SeldomSeenKid wrote:As a rule of thumb your subject should never be in front of the projection-plane for numerous reasons, thus your subtitles can live on the projection plane without conflicting with your subject. Only objects, which are not cropped by the screen borders can stay out of the screen without causing eye-strain.

This is an old vision of 3D. This is what almost killed 3D recently. Rules made by thoricians who don't make films. Every time someone said to me "it will not work in 3D", I tried to do it and showed it to them.
Every time I git the same "oh, ok, it works, you were right". 3D is not just maths. It's also the artistic way you do it.
Believe me. I directed and stereographed more than 500 hours of 3D contents in the last 15 years and also wrote a book about 3D filming. (https://www.amazon.com/Tourner-en-3D-re ... 134&sr=8-1)

You talk about the 3D in Avatar 2. There are many objects cropped in Avatar 2 and no one cares. And yes, viewers want "out of the screen" 3D.

Wolfgang Woehl wrote:T
Simply sticking subtitles to the screen plane does not work at all.

Exactly.
We have to be able to place the subtitles in a 3D space, not only on ne screen plane and export them with XYZ format
Sometimes the subtitles have to be on the left or on the right, but also in front of the screen. This is what is missing in DVR.

Fabien Remblier wrote:This is an old vision of 3D. This is what almost killed 3D recently. Rules made by thoricians who don't make films. Every time someone said to me "it will not work in 3D", I tried to do it and showed it to them.
Every time I git the same "oh, ok, it works, you were right". 3D is not just maths. It's also the artistic way you do it.

I am sure this is not what killed 3D in cinemas. There is no old and new 3D - there's only correct 3D. Surely as an artist you can bend the rules but you cannot break them.
As I stated, I never made a 3D movie but I've seen enough that where just crap in terms how 3D was used. This is where artistic freedom kills 3D.

SeldomSeenKid wrote:

Fabien Remblier wrote:This is an old vision of 3D. This is what almost killed 3D recently. Rules made by thoricians who don't make films. Every time someone said to me "it will not work in 3D", I tried to do it and showed it to them.
Every time I git the same "oh, ok, it works, you were right". 3D is not just maths. It's also the artistic way you do it.

I am sure this is not what killed 3D in cinemas. There is no old and new 3D - there's only correct 3D. Surely as an artist you can bend the rules but you cannot break them.
As I stated, I never made a 3D movie but I've seen enough that where just crap in terms how 3D was used. This is where artistic freedom kills 3D.

Agree to disagree !

In the meantime I made a small script to at least position all 2D subtitles in a safe position in front of the content. This does not really solve all the problems, but help people who are in despair with 2D subtitles in their 3D movie.

If someone would like to test it I would be very happy for feed back.

https://github.com/stereo3d/DCDM_3DSUB_XSLT

I'm in !

I had the chance to test the xsl transformation to convert 2D subtitles in 3D space as described above.
The test worked as expected. Here some screenshots for the test screening.



main compliance test picture at 4K60fps



subtitles placed over real picture at 1% negative parallax.

it looks great.

while preparing a new edition of the 3D program of the Busan International Short Film Festival, I developed this plugin to facilitate the creation of 3D DC subtitles.



I share the result with this group. You can find a detailed video tutorial on the workflow here:



The code is on Github:
https://github.com/stereo3d/StereoscopicSubtitlePlugin

I owe a lot to the experience and work of many in this forum. So I also want to give back something.
This has so far solved the problem for me. It is far for perfect. I learned a lot about Resolve on the way. It took more time than expected, because the API was changed in version 19.

I hope this can help other festivals or people who are making 3D movies.
I wish it I can see your movies at the next festival.

I would be glad if someone likes to try it out and let me know if this is helpful.
Please be aware this is a very early version.

Good stuff! And thanks for sharing ideas and code.

Two little things wrt workflow:

- Resolve's subtitle tracks are still broken wrt italics (usually using SubtitleEdit to render images instead)

- for some reason Resolve doesn't link convergence for 3D assets by default (the little chain icon next to the convergence field). So that 's 600 more clicks for a feature length subtitling project lol

SeldomSeenKid wrote:So do I. Say you want to place your 2D object (text) a given distance X in front OR behind the screen-plane. The projection offset between the two (left/right) representations of this object will be equal in both cases. As a consequence your brain will not be able to determine the correct depth.

Old but worth addressing. Offset is just one aspect, other aspect is where the text is placed in left and right eye images (there are two variants with same offset, AB and BA). This is what determines whether it appears in front or behind screen plane.

@Wolfang Woehl and @Hendrik Proosa
thanks for feedback and additional input.

Suggesting a simplification of the workflow, where you can skip the entire 1st part (rendering and re-splitting the subtitle track):

1. open your subtitles in SubtitleEdit
2. export "Final Cut Pro + image..." (use full image path, set full frame image, set dimensions, use PNG etc.)
3. Resolve > File > Import > Timeline to import that FCP timeline with all subtitle images as 1st class citizens
4. select all subtitle images > Convert to Stereo

For the actual positioning pass, on the Color page, consider disabling autoselect for non-subtitle tracks (V1 etc.). This will then always autoselect the current subtitle image - and not the other/image assets.

Wolfgang Woehl wrote:Suggesting a simplification of the workflow, where you can skip the entire 1st part (rendering and re-splitting the subtitle track):

1. open your subtitles in SubtitleEdit
2. export "Final Cut Pro + image..." (use full image path, set full frame image, set dimensions, use PNG etc.)
3. Resolve > File > Import > Timeline to import that FCP timeline with all subtitle images as 1st class citizens
4. select all subtitle images > Convert to Stereo

For the actual positioning pass, on the Color page, consider disabling autoselect for non-subtitle tracks (V1 etc.). This will then always autoselect the current subtitle image - and not the other/image assets.

That is indeed a very good idea.