Hi Maria,

A similar situation was recently posted here: https://my.aximmetry.com/post/4588-i-have-a-ptz-camera-and-id-like-to-try-v
I think the information there might also apply to your case of recording in portrait.
If your situation is the same, feel free to contribute to that thread if my current solution doesn't fully resolve your issue. If your situation differs, could you please describe it in more detail?

Warmest regards,

As I understand it, we are talking about a full portrait output here (in the other post), I need to keep the horizontal output and change only the orientation of the camera, not the world.

I still want to shoot the scene with 1920x1080, full portrait scene not needed in my case, just to flip tracked camera.

In my case, there is simply no sensor on the camera that could transmit the orientation after calibration. If you use a virtual camera - there is a button to change the orientation, unfortunately, there is no such button with tracked camera.

And when i try to flip camera/world - they start spinning each other)

But i think i can try to set scene to 1080x1920 and rotate it backwords)) Changing Delta Cam Transf too ))

Will try it today, thank you!

With best regards, Maria.

Hi Maria,

Unfortunately, the solution in my linked post won't work for your situation because you need the final image to be in landscape (resolution of 1920x1080).

Note that this is challenging because the lens distortion and tracking calibration must align with the aspect ratio of both the video input and the rendering frame size. For correct tracking and distortion alignment, the same aspect ratio throughout the composition must be maintained.
The solution in my linked post works because the landscape aspect ratio is maintained throughout the composition.
In theory, you could rotate the scene by 90 degrees and then roll the camera using a Camera VR path or place the input on a landscape video inside the camera compound before it is processed by anything else. However, in these cases, you would lose the image quality of the talent that you achieve when recording in portrait mode rather than landscape. Additionally, for example, the lens distortion would be incorrect in such a rotated setup.

I recommend using a virtual camera along with the CROP panel's Portrait parameter to rotate the input video. Is there a specific reason you're hesitant to use a virtual camera if active tracking is unnecessary? There are likely ways to replicate any features you find missing from the virtual camera.

If you choose to continue using a tracked camera setup, you should not rotate your camera sideways. Instead, use the STUDIO panel to remove non-green parts of the shot. More details can be found here: https://aximmetry.com/learn/virtual-production-workflow/green-screen-production/tracked-camera-workflow/studio-control-panel/

I just don't fit into the camera completely when it's not vertical ( Too little distance.

It would be a little better if the lens was 16mm instead of 23, but not much, it would probably cut off my legs.

I would like to move around the scene while maintaining proportions, for this I need tracking. In this regard, the virtual camera is not suitable. I used it before) I would like to raise the quality to a higher level and it is very difficult to maintain the position in height, since all scenes are different, on a virtual camera. Too many scenes in live broadcast.

I don't have problems with cropping the frame for greenscreen, thank you! (and i use Studio Panel too))

With best regards, Maria.

Hi,

You previously suggested you wouldn't use tracking for camera movement, but now you are suggesting you will. In this case, you are right; the virtual camera cannot replace the tracked camera.

Otherwise, you could create a logic to maintain a consistent distance from the camera to the billboard across your projects, similar to the Put In Front function. If you're interested, I can guide you through how to make it:

With the tracked camera compound, you can consider rotating and then placing the input on a landscape video within the camera compound before any other processing occurs. I would still advise against it. Instead, for example, consider using a camera that records in portrait resolution.
Anyway, to achieve this, open the camera compound using Edit Compound In-Place and double-click to enter it:

You can learn more about Edit Compound In-Place and Linked compound here: https://aximmetry.com/learn/virtual-production-workflow/scripting-in-aximmetry/flow-editor/compound/#linked-compound 

Inside the compound, repeat this process for the INPUT %N= compound you are currently using:

In it, enter the Line %N compound:

Between the Vignette and TrkCam_Record compounds, you rotate the video and then fit it into a landscape resolution.
First, rotate it using the Rotator module. Then, with the Placer module, fit it into a landscape-sized video by adjusting the Out Size pin to maintain a 1920x1080 (16:9) aspect ratio. To preserve the video quality, you should maintain the 1920 height. This means the best if the Out Size resolution is set to 3414x1920 (16:9 resolution and 1920 height). Set the Fit Mode to Keep to prevent any resizing. Additionally, set the Filter to Clamp to avoid 0 alpha in unfilled video areas, as the keyer can not work with 0 alpha.

Finally, you must turn off Lens Distortion, as the Distortion data won't be rotated:

Warmest regards,

"Otherwise, you could create a logic to maintain a consistent distance from the camera to the billboard across your projects, similar to the Put In Front function. If you're interested, I can guide you through how to make it:"

Yes, i`m interested! Thank you very much in advance! 

I'm afraid it won't help me though) I don't know where exactly to stand and the point keeps shifting ( Since I don't always stream on a green background and don't always stand in this place, I periodically go there during the stream. In general, most likely I will have to somehow solve the issue with the horizontal mode (

As I wrote before, I don't plan to move around with the camera in my hands while working live, so I don't need to use the Mars that I have.

I would like to track the movement of the legs inside the scene so that there are no height differences on the floor, when i`m moving at least forward and backward. 

Maybe I don't need it for that? But the proportions will still change if i understand right) (or not? ))

"Instead, for example, consider using a camera that records in portrait resolution." 

Sadly, i`m using it live with capture card, as far as i know, capture cards can't make it 1080x1920.

If it wasn't live - everything would be good)

Thank you very much for the second part too! I will also try to use this method, but in such resolutions even with 4090 - I am afraid there will be problems with some scenes on live.

I would like to achive something like this https://www.youtube.com/watch?v=6zyfpP1PelA, with htc vive and in portrait mode of the camera on live stream) 

But so far from what I understood - it is very difficult to do this with portrait mode. Therefore, I will try to somehow solve this issue without it i guess.

With best regards, Maria.

@Maria Moon: I think there is some fundamental misunderstanding about how camera tracking vs. virtual cameras when it comes to scale. Why would you need a tracked camera to maintain correct scale if the camera is stationary? Just set the virtual camera at the correct height, matching the height of the physical camera. Then match the field of view and the distance to the physical distance between you and the camera. This should give identical results to a locked-off, tracked camera. 

Note that if you move backwards or forwards on the green floor, you will still become detached from any virtual shadows and reflections, even if you’re using a tracked camera. The video signal is still flat 2D, so there’s no way to cast accurate reflections or shadows in 3D space without a tracker on the talent. Even then, the shadows and reflections won’t look completely realistic. 

Hi Maria,

Everything Johan mentioned is accurate and worth considering.

About the Put In Front function:
First you should locate the Put In Front's modules by opening the compound located at
[Common_Studio]:Camera\VirtualCam_Unreal\Elements\VCamU_Billboard.xcomp
Within this, enter the Put In Front compound:

Here, the Delta pin of the Translation Z module sets the distance (in meters) by which the billboard is positioned in front. You should set this to the distance between your camera and your talent (billboard) in your real-world studio.
You could copy, and with some modifications, you could use this whole logic in your project compound. However, it is easier to modify this Billboard compound by exposing the Delta pin on the Billboard's control board.
To do so, expose the Delta pin to the left within the compound:

Exit the Put In Front compound by pressing CTRL + Back.
Expose the Delta pin again to the left.
Save the compound. The Delta pin will then be visible on the Billboard's panel for the virtual Unreal cameras.
If needed, use the Reorder Input Pin... option to position the Delta pin just below the Put In Front pin.

Note that whenever you update Aximmetry, the VCamU_Billboard.xcomp file will be overwritten, requiring you to repeat the changes outlined above after each update.
Also, you should ensure your virtual scenes have the floor set at 0 height. The current "Put In Front" logic places the billboard directly on the 0 height if it is close to 0 height.

About the second part of editing the tracked camera compound.

The 3414x1920 is just the video resolution, not the rendering resolution. 4090 will have an easy time handling that resolution as video and texture in the virtual world.

And this video https://www.youtube.com/watch?v=6zyfpP1PelA has lots of tracked camera movements for which you will have to use your Vive tracker. Also, it mixes the virtual world with the real world's studio image, for example, here:

You can quite easily do this even with a tracked camera using the new Window Mask Mode: https://aximmetry.com/learn/virtual-production-workflow/green-screen-production/tracked-camera-workflow/studio-control-panel/#mask-modes
With a static virtual camera, you just need to blend it over the final render of your camera picture.
However, in both cases, virtual camera paths will not work on the studio's image for obvious reasons.

Warmest regards,