Hi SGO team!

Happy New Year !

I'd like to know how Boutique works when we apply VR View and VR Stitch to map camera fisheyes to LatLong (equirectangular) or Cube (cross or EAC). As you know, pixel density is not the same and an equirectangular rendering is more destructive than cubemap. For a new job, I need to work with cubema, but all my work was made on a a latlong mapping (masks, VR View Yaw/Pitch/Roll alignments, stabilisation, etc...). So, I can't just change latlong to cubemap on my first VR Stitch effect (mapping camera rushes). All the geometry modifications beyond couldn't work anymore!

If I apply a cubemap at the end, respecting the pixel ratios (LatLong 16384x8192 -> Cube4x3 16384x12288), it seems I get the same definition when I check one face (4096x4096) than if I map firstable camera rushes in Cube 4x3... That means Boutique don't bake / burn pixels between a LatLong to Cubemap mapping with a VR View, wherever we use it ?

I just wanted to be sure about that...

Maxime - Paris

Hi Maxime,

Latlong is indeed a somewhat wasteful format; about 1/3 of the pixels are lost due to the high horizontal resolution at the poles. That said, it doesn't harm the image quality. Cubemaps are more efficient, but they can be tricky to use as a working format because of edge interpolation issues. Using them for final delivery, as you’ve done, is definitely the right call. Your current pipeline—working in latlong and delivering in cube—is the standard and safest way to go.

As for the baking process, there is indeed a remapping of pixels using subpixel positions during the conversion. Mistika handles this via bicubic filtering, which, as you noted, shouldn't cause any noticeable degradation provided the 'pixels per degree' resolution remains consistent. You might see some minor impact on extremely fine details (like converging lines on a test chart), but you'll rarely encounter those in real-world footage.

Thanks for your answer !