While Video AI can handle 360° videos, our models aren’t primarily trained on equirectangular footage. Thus, we can’t assure full compatibility with VR inputs or preservation of 3D alignment with 3D SBS inputs. While some users have had success with these formats, they’re not officially supported yet. You may need to re-inject 360° metadata into the processed output for optimal results.
In regards to model selection, I’d recommend looking through our Docs page which includes descriptions of each model’s use case. You may want to start with Artemis.