Haha ya quite a few of them, but that’s is so computationally taxing I’m running 200 3k exports for automated colmap data sets from a input hotfolder with a Python subtle sharpen and grain process on the backend - it takes about 25 min per image to full data set - but the current limitation is the SLat texturing that works great bc no UVs and projects PBR textures onto a 16 mill poly space in latent space which is great but without month of reconditioning and training (and NOT on a 5090) that’s the biggest limitation atm. Post upscale and enhancements are notable for sure as a phase 2.