Next Render Tile Button
This idea has been lingering in my thoughts for a long time now and it's great that we finally have a hub to post these in
It would be nice to have like a "next" tile button in the render window so you don't have to wait for the complete samples to finish on a certain tile before proceeding to the next one—this saves time, on a supervised rendered basis.Say, you are rendering an 800 x 800 image using 1000 samples and will be rendered in a total of 4 tiles. On the 1st tile, the rendered image is already looking good and noise-free at only 300 samples. Instead of waiting for it to reach 1000 samples which would take relatively some time, you can press the "next" button to proceed to the 2nd tile, and so on.
But I only see this being useful when using GPU rendering which uses large chunks of tiles by default.
I'd love to hear your thoughts on this. :)
Reyn
5 Comments
Disagree too. Better wait for adaptive sampling. This could be though for quickly see the whole render. Skipped tiles should be put in queue and be calculated later. But again, while supervising, better a "mouse aware" tile order
V-ray has a similar feature and from what I can see from other students, it's very helpful because you can manually check specific parts of the image first, then the renderer continues with the rest. If you don't like the result of that area, just cancel and rework like normal.
It may be helpful to included a mock-up of where this Next Tile button would go in the interface.
I disagree with this for the fact that tiles would be almost guaranteed to be inconsistent. Even if you see it as nearly noise-free, there is still the fact that the renderer has given each tile different sample numbers, leaving potential for a visible tile border (I have seen one or two samples make a difference). This would be great for doing tests for lighting, color or maybe even post, but not for a final render. This is what a test render is for, however, so I see no real benefit. Plus, you wouldn't have this feature for terminal rendering, nor could you easily implement it in a python script; you would be stuck with a fixed sample count anyway.
Perhaps there could be a way to test average variance between pixels, such that if a general area of pixels falls below a custom variance threshold (say, luminance or chroma difference of <2%) it would move to a new area. I believe this might fall under the category of adaptive sampling, however, which has its own thread already.
Thanks for the thoughts. Yes, the average variance would be a better method, I believe. Thanks a lot.