There are many options for reducing render times but they all come at the cost of quality / accuracy. For example:
- reducing output resolution
- lowering render sample counts
- lowering light path bounces
- reducing polycounts of models
- reducing the objects in the scene
- reducing number of hairs
- reducing numbers of particle…
on and on….it’s very scene specific but the general theme is reduction. For this scene you could try the first three since the others aren’t relevant.
Maybe this isn’t you, but a lot of people ask this question hoping there’s a way to get the same high quality, highly accurate render in less time. Which isn’t a thing. It’s all compromise.
For what it’s worth, 3 min render time isn’t bad. I’m usually content with 10 minutes or less per frame (so long as it fits the complexity of the scene). I.e if the default cube takes 12 min to render, something is wrong and needs to be troubleshot, whereas a realistic character taking 12 minutes makes sense. It’s usually when I get into 20+ minute render times that I start to seriously sacrifice quality/accuracy for optimizations.
