If crappy phone photos are enough to make a decent depth map, I think Stablediffusion can do this today. The only missing piece would be for the band to train

If crappy phone photos are enough to make a decent depth map, I think Stablediffusion can do this today. The only missing piece would be for the band to train a model on high quality photos from specifically their concerts. ControlNet is really neat