Quick comparison of an old dataset from postshot compared to Spirula.
Video capture using a phone. S24.
Figures on the left are processed with Image Alignment in RealityScan then 3DGS trained in postshot.
Quick video of process here: https://youtu.be/oYezeGdum_Q
The figures on the right are processed in Spirula Studio (2026.10.6).
Just drag and dropped the video, ticked depth estimation and everything else left as default. so it aligned and processed the splat.
I think spirula produces a better result here.
For fairness, I should repeat the test with the latest Postshot, and also let it select and align the images from the video. But don't have a paid licence atm, so cant export etc.
Edit: Have just ran the same video in postshot v1.1.69 (2026-08-18) and it has produced a much better result. so either the reality scan step provided a poor dataset for postshot. or postshots training has got better. its camera alignment is certainly much much faster - as this was the reason to use reality scan in the first place! Wouldn't bother with reality scan unless you needed it for control or scale etc. Postshot did still create quite a lot of floaters though.