Chiesa

I just used two videos (~2-3 mins long each) captured with a drone flying around the subject at different angles. Then imported them straight into Postshot and, by using 1000 frames out of ~20000, I let the software do the job. The only manual thing was to crop the region directly inside Postshot. Pretty straightforward workflow and quite impressive final result considering it took 10 mins to capture and ~3 hours to process

I assume this has been posed with reality capture and there is even a clean polygon-model of it. Thus you had perfect masks for training the splats. Is that assumption correct?