AA and elephant foot



Nice software, a few questions turned up during evaluation:

  1. Anti-Aliasing (AA) is applied after slicing to each individual layer. That means there is no AA in z-direction?
  2. Applying AA to only a few surfaces instead of the entire table (or part) is not possible? I’m not aware that any slicer can do this but it would be certainly very helpful for our parts.
  3. We want to grow some parts directly from the table, mainly because several planes in the part need to be coplanar to the table including the plane attached to the table. Is there an “elephant foot compensation”? Prusa Slicer seems to have this option for SL1 though I don’t know whether it really works for SL1.

Thanks for some hints,


Hi Marc,

  1. Correct, AA is applied on the GPU per layer. Currently there is no AA in Z-direction.
  2. This is currently not possible, it might be doable in CPU slicing mode but to be honest this is the first time I’ve heard this requested.
  3. You can correct for this using XY-offsets. (print profile -> slice post processing) This can be defined for the bottom layers separately. It does require some tuning though, as this is strongly material (transparancy) and build table (reflectivity) dependant.




Hi Jan,

Thanks for the quick response. Didn’t know 3), will try it.



I had a quick look at some slices with and without AA (4) applied and there seemed to be no difference between them. I need to look at this more carefully, but as this thread popped up I thought I’d mention it.


Hi nzfinescale,

Depends i guess totally on how it’s sliced and what is sliced.

slicing on:
GPU you should be able to see a clear difference.
CPU: it’s activated by default by the 2D drawing library at the moment.
So the AA settings is only for GPU slicing as we speak.



So even though you select AA, you may or may not get it and there is no warning or notice either way? If so, that doesn’t sound very good.


So the extent of AA is supposed to be determined by the AA setting in the ‘slice post-processing’ tab of the print profile. It would seem to also be affected by the ‘force slice generation in the CPU’ check box in the machine ‘slice setup’.

Attached image shows the results I got for the same slice of a test piece using different settings.

So with the CPU forced off (so GPU slicing), there is an expected effect of the AA (top row).

With CPU slicing (bottom pair) you get AA regardless of AA setting and the result appears to be somewhat more smoothed than the 8x sliced with the GPU. This behaviour doesn’t seem to be intuitive or documented.

I’ve been following threads on the Phrozen forum looking at smoothing. There has been some interesting stuff involving adding Gaussian blur in Photoshop, and adding diffusers below the LCD (To reduce lens artifacts, but smoothing was a side effect). It seems therefore that a good AA solution would be the best way to smooth prints, but I’m not sure the current method offers all that is possible. Still theoretical on my part as my Phrozen has yet to be delivered.



Yes the behaviour you see is correct at this point in time.

GPU slicing -> it applies real AA as it should be. 2x, 4x or 8x. All implemented on your GPU.
(background here: https://www.formware.co/article/AntiAliasing )

The CPU slicing was introduced 2 months back.
Currently we use C# GDI+ (graphics.drawing) implementation to draw the images after the CPU calculates the slices (polygons). It’s not defined that this drawing lines/polygon’s actually gets GPU acceleration (and thus AA support). Sometimes it will, sometimes it won’t. See this thread in case you are interested: https://stackoverflow.com/questions/31541636/does-c-sharp-natively-use-gpu-for-graphics
and then Hans Passant’s comments.

As per default we’ve coded it to smoothingmode==antialias, which is used if you get GPU support in 2D.
I think it would be wise to turn it off as well; however i didnt’ test yet how rough the lines will be then…
Currently this would only be an issue for the non-AA machines out there…

Regarding smoothing. I’m aware of the gaussian blur.
To be more precise (and realizing we should be ;)) it is used currently when you are using GPU slicing + AA + XY Offset correction to bring back the AA effect after offsetting the pixels with a rolling ball principle. When you offset a pixel with a rolling ball you loose the AA effect. Haven’t found a good way of getting it back as you are either creating or removing geometry in 2d.
(will need to put that in the documentation btw… ).



very interesting points, thanks. Figure 1 in https://www.formware.co/article/AntiAliasing seems to show AA in z-direction. Maybe you could mention in the post that Formware currently only supports AA in xy direction (or add AA that considers voxels :slight_smile: )


Yes you are right… unlucky pick of image.